 

 

I .71.} I 0
:63: 13:3.

:9

 

“5.11.. v.

 

ext.” , .3

5:... .2
... w: "in.
g .

a»?

t

z , ..
:35. «a? r

.. 3.7. .2?
i. z...

.53;

35...:

. :.

1.3.3..
it. :13?!
V .. .

x.“ “as

I .3.

 

 

 

 

 

..ﬁaﬂ2...¢x: 7:". u :1,’ u. 1.51. hr
.. . a. .
. ; éiéé E

.LX»

 

 

 

  

 

 

. 1..

1%

l 5.:

 

in...“ a .

étmmm

 

31

k.

43%

m.

_

2..

:23. .

2

u‘

c 1......

a.

1 2%... -
.. new“... an: iv
I l‘ﬁa ,

 

.nILM

.
$3.5.
.

J
v )5
(rd . 0.1.53
1‘

.lu
i

3

3‘

5

«iv
IF:

iLw.
1‘ .

I.

ma
3.
‘5.ﬁu.:
a”,
x . p.15.

. ..

   

1. ‘3

 

LIBRARIES

LI MICHIGAN STATE UNIVERSITY
p , EAST LANSING, MICH 48824-1048
cz,/‘r
a. -
.43 / 6? I J2. 3

This is to certify that the
dissertation entitled

THE USE OF LONG TERM PAVEMENT PERFORMANCE
DATA FOR QUANTIFYING THE RELATIVE EFFECTS OF
STRUCTURAL AND ENVIRONMENTAL FACTORS ON THE
RESPONSE AND PERFORMANCE OF NEW FLEXIBLE
PAVEMENTS

presented by

Syed Waqar Haider

has been accepted towards fulﬁllment
of the requirements for the

Ph. D. degree In Civil Engineering

 

 

AMA

Major Professor’sSTQ'W

3/14/?a15‘

‘Date

MSU is an Affirmative Action/Equal Opportunity Institution

 

PLACE IN RETURN BOX to remove this checkout from your record.
to AVOID FINES return on or before date due.
MAY BE RECALLED with earlier due date if requested.

 

DATE DUE DATE DUE DATE DUE

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

2/05 chTRC/DateDueJndd-p. 15

 

THE USE OF LONG TERM PAVEMENT PERFORMANCE DATA FOR
QUANTIFYING THE RELATIVE EFFECTS OF STRUCTURAL AND
ENVIRONMENTAL FACTORS ON THE RESPONSE AND
PERFORMANCE OF NEW FLEXIBLE PAVEMENTS

By

Syed Waqar Haider

A DISSERTATION

Submitted to
Michigan State University
in partial fulfillment of requirements
for the degree of

DOCTOR OF PHILOSOPHY

Department of Civil and Environmental Engineering

2005

ABSTRACT

THE USE OF LONG TERM PAVEMENT PERFORMANCE DATA FOR
QUANTIFYING THE RELATIVE EFFECTS OF STRUCTURAL AND
ENVIRONMENTAL FACTORS ON THE RESPONSE AND PERFORMANCE OF
NEW FLEXIBLE PAVEMENTS
By

Syed Waqar Haider

Considerable progress has been made over the past 50 years in the ﬁeld of
pavement engineering. While much has been learnt about designing and maintaining
pavements economically, still there exists a need for improving existing procedures to
meet increasing infrastructure needs with limited resources. At present, highway
agencies lack adequate information on the inﬂuence of pavement drainage on the
performance of ﬂexible pavements. There is very limited experimental ﬁeld data to.
quantify its inﬂuence or effect on the pavement performance. The interactions of the
structural factors with other variables such as climate are not well understood and
therefore, still needs to be explored in order to improve relationships between pavement
performance and response. As a result, controlled ﬁeld experiments are necessary to
answer questions which are currently not explained by the theoretical modeling.

This research documents and presents the results of relative inﬂuence of design
and construction features on the response and performance of in-service new ﬂexible
pavements, included in SPS-l and SPS-8 experiments of the Long Term Performance
Pavement (LTPP) program.

In summary, for pavements in the SPS-l experiment, base type seems to be the
most critical design factor for fatigue cracking, roughness (IRI), and longitudinal

cracking-WP. This is not to say that the effect of HMA surface thickness is not

Syed Waqar Haider

Signiﬁcant. In fact, the effect of base type should be interpreted in light of the fact that a
dense graded asphalt treated base effectively means thicker HMA layer. Drainage and
base type, when combined also play an important role in improving ﬂexible pavement
performance, especially in terms of fatigue and longitudinal cracking. Base thickness has
secondary effects on performance, particularly in the case of roughness and rutting.

Subgrade soil type seems to be playing an important role in ﬂexible pavement
performance. In general, pavements built on ﬁne-grained soils have Shown worst
performance, especially in the case of roughness. Also, climate is a critical factor in
determining ﬂexible pavement performance. Longitudinal cracking-WP and transverse
cracking seems to be associated with Wet Freeze environment, while longitudinal
cracking-NW? seems to be the dominant in “freeze” climate.

On average, for pavements in the SPS-8 experiment, those located in WF zone
have more fatigue cracking, longitudinal cracking-NWP, and roughness than pavements
in other climates. Also, in general, pavements constructed on “active” subgrade (frost
susceptible or expansive) soils have more longitudinal cracking-NWP, transverse
cracking, and fatigue cracking than pavements on “non-active” soils. Pavements located
in “wet” climate, on average, have higher change in IRI than those in “dry” climate.
Furthermore, pavements located in WF zone and those built on active soils have higher
change in IRI.

Although most of the ﬁndings from this research support the existing
understanding of pavement performance, the methodology in this study provides a
systematic outline of the interactions between design and site factors as well as new

insights on various design options.

DEDICATION

This work is dedicated to my parents, late younger brother Syed Ali Haider, late uncle
Syed Nazir Hussain Mashadi and my family, without whom this work would not have
been achievable. Their emotional support and prayers consistently provided the

motivation and inspiration to achieve this goal.

iv

ACKNOWLEDGEMENTS

I would ﬁrst like to express my gratitude to Dr. Karim Chatti for his support,
ideas and guidance throughout my research period at Michigan State University. I would
also like to extent my appreciations to Dr. Neeraj Bush and Dr. Gilbert Baladi for their
valuable advices.

The assistance and guidance provided by Dr. Richard Lyles, Dr. Bruce Pigozzi
and Dr. Denis Gilliland for statistical analysis have played an important role in the
completion of this research study. I would like to convey my sincere thanks and prayers
to all the above professors at Michigan State University.

My appreciation is extended to the NCHRP for sponsoring this research work and
providing ﬁnancial support during the period of this research. I would also like to extend
my thanks to my fellow graduate students Mr. Aswani Pulipaka, Mr. Hyungsuk Lee and
Mr. Hassan Salama for their help and assistance during the period of this research.

Finally, I would like to appreciate the support and effort of my wife who took care

of our three children Azam, Raza and Zanaib during this time.

TABLE OF CONTENTS

LIST OF TABLES ...........................................................................

LIST OF FIGURES .........................................................................

CHAPTER 1 INTRODUCTION

1.1 INTRODUCTION ..................................................................
1.2 PROBLEM STATEMENT .........................................................
1.3 RESERACH OBJECTIVES .......................................................
1.4 SCOPE OF STUDY ................................................................

1.5 REPORT ORGANIZATION ......................................................

CHAPTER 2 LITERATURE REVIEW

2.1
2.2
' 2.3

2.4

GENERAL ...........................................................................
BACKGROUND ....................................................................
SUMMARY OF RELATED LTPP STUDIES .................................
2.3.1 Factors Affecting Flexible Pavement Performance ....................
2.3.2 Relationship between Pavement Response and Performance ........

REVIEW OF STATISTICAL METHODS ......................................

CHAPTER 3 DESCRIPTION OF SPS-l AND SPS—S EXPERIMENTS

3.1
3.2

3.3

3.4

INTRODUCTION ..................................................................
STRATEGIC STUDY OF STRUCTURAL FACTORS FOR FLEXIBLE

PAVEMENTS—SPS-l ............................................................
3.2.1 Experiment Design .........................................................
3.2.2 Discussion on the SPS-l Experiment Design ...........................
3.2.3 Current Status of the Experiment (Release 17 of DataPave) . . . .
3.2.4 Construction Guidelines for SPS-l Experiment ........................

STRATEGIC STUDY OF ENVIRONMENTAL FACTORS FOR

FLEXIBLE PAVEMENTS—SPS-8 EXPERIMENT ..........................
3.3.1 Discussion on the SPS-8 Experiment Design ..........................
3.3.2 Current Status of the SPS-8 Flexible Pavements .......................
3.3.3 Construction Guidelines for SPS-8 Flexible Pavements ..............

INSTRUMENTED SPS TEST SECTIONS .....................................
3.4.1 SPS-l Experiment Sections ...............................................

CHAPTER 4 DATA AVAILABILITY AND EXTENT

4.1
4.2

INTRODUCTION ..................................................................
IDENTIFICATION OF DATA ELEMENTS ...................................

Vi

 

x
xiii

1
l
3
4
5

6
7
8
9
17

21

25

25
26
32
33
36

45
46
46
46

49
50

52
52

4.3

4.4

DATA AVAILABILITY IN SPS-l EXPERIMENT ........................... 59
4.3.1 General Site Information .................................................. 60
4.3.2 Material Data ............................................................... 65
4.3.3 Design versus Actual Construction Review ............................ 65
4.3.4 Extent and Occurrence of Distresses .................................... 75
4.3.5 Dynamic Load Response Data (DLR) —— Flexible Pavements ...... 92
DATA AVAILABILITY IN SPS-8 EXPERIMENT — FLEXIBLE
PAVEMENTS ....................................................................... 95
4.4.1 General Site Information .................................................. 95
4.4.2 Material Data ................................................................ 97
4.4.3 Design versus Actual Construction Review ............................ 100
4.4.4 Extent and Occurrence of Distress ....................................... 101

CHAPTER 5 ANALYSIS METHODS

5.1
5.2
5.3

5.4
5.5

INTRODUCTION .................................................................. 106
PERFORMANCE INDICATORS ................................................ 106
OVERALL STATISTICAL ANALYSIS METHODS ........................ 115
5.3.1 Analysisofvariance(ANOVA).......................... ................ 119
5.3.2 Discussion on Analysis of Variance (ANOVA) ....................... 141
5.3.3 Comparison of Means (One-way AN OVA) ............................ 142
5.3.4 Extent of distress ........................................................... 145
5.3.5 Linear Discriminant Analysis ............................................. 145
5.3.6 Binary Logistic Regression ................................................ 145
SITE-LEVEL ANALYSIS METHODS .......................................... 146
METHODS FOR INVESTIGATIN G APPARENT RELATIONSHIP

BETWEEN PAVEMENT RESPONSE AND PERFORMANCE ............ 151
5.5. 1 Univariate Analysis ......................................................... 15 1
5.5.2 Multiple Regression Analysis ............................................. 152

CHAPTER 6 ANALYSIS RESULTS FOR SPS-l EXPERIMENT

6.1
6.2

6.3

6.4
6.5

INTRODUCTION .................................................................. 153
PREVIOUS STUDIES .............................................................. 154
6.2.1 Summary of Findings ...................................................... 154
EFFECT OF CONSTRUCTION ON PAVEMENT PERFORMANCE 158
6.3.1 Construction-related Issues ................................................ 159
6.3.2 Drainage-related issues .................................................... 173
SPS-l PROJECT PERFORMANCE SUMMARIES ........................... 175
SITE-LEVEL ANALYSIS ......................................................... 183
6.5.1 Effects of design features on performance — Paired Comparisons at
Level-A ...................................................................... 186
vii

6.6

6.7

6.8

6.5.2 Effects of design features — Paired Comparisons at Level-B ......... 191

OVERALL ANALYSIS ............................................................ 193
6.6.1 Extent of Distress by Experimental Factor .............................. 193
6.6.2 Frequency-based methods ................................................. 205
6.6.3 Analysis of Variance ....................................................... 211
6.6.4 Effect of Experimental Factors on Pavement Response ............... 259

APPARENT RELATIONSHIP BETWEEN RESPONSE AND

PERFORMANCE ................................................................... 264
6.7.1 Overall Analysis—Explanatory Relationships ......................... 264
6.7.2 Site Level Analyses— Predictive Relationships ........................ 266
6.7.3 Overall Analyses— Predictive Relationships ........................... 274
6.7.4 Dynamic Load Response for OH (3 9) test sections .................... 284
SYNTHESIS OF RESULTS FROM ANALYSES ............................. 288

CHAPTER 7 ANALYSIS OF SPS-8 EXPERIMENT

7.1
7.2

7.3

INTRODUCTION .................................................................. 306
EFFECTS OF ENVIRONMENTAL FACTORS 1N SPS-8 EXPERIMENT
FOR FLEXIBLE PAVEMENTS .................................................. 307
7.2.1 Site-Level Analysis ......................................................... 307
7.2.2 Overall Analysis ............................................................. 308
7.2.3 Comparison between Similar Designs of SPS-8 and SPS-l
Experiments ................................................................. 3 14
SUMMARY OF RESULTS ........................................................ 316
7.3.1 Effect of SPS-8 Experimental Factors on Performance of Flexible
Pavements ................................................................... 316

CHAPTER 8 CONLUSIONS AND RECOMMENTATIONS

8.1
8.2

8.3

8.4

INTRODUCTION .................................................................. 3 1 8
EFFECTS OF STRUCTURAL FACTORS FOR FLEXIBLE

PAVEMENTS — SPS-l EXPERIMENT ....................................... 320
8.2.1 Effect of Design and Site Factors on Pavement Performance ........ 320
8.2.2 Effect of Design and Site Factors on Pavement Response 329
8.2.3 Apparent Relationship between Response and Performance ......... 330

EFFECTS OF THE ENVIRONMENT IN THE ABSENCE OF HEAVY
TRAFFIC FOR FLEXIBLE PAVEMENTS —— SPS-8 EXPERIMENT 332

LIMITATIONS OF THE EXPERIMENTS AND ANALYSES .............. 333

8.4.1 Experiment-related issues ................................................. 333

8.4.2 Data-related issues ......................................................... 334
viii

8.5 RECOMMENDATIONS FOR FUTURE DATA COLLECTION AND
RESEARCH ..........................................................................

REFERENCES ..............................................................................

ix

 

335

339

A

‘-

LIST OF TABLES

Table 3-1 Full Factorial Experiment Design Matrix ......................................................... 29
Table 3-2 Fractional Factorial Experiment Design Matrix and Replications ................... 30
Table 3-3 SPS-l Experiment Design Matrix .................................................................... 31
Table 3-4 SPS-l site factorial -— From DataPave 3.0 ...................................................... 35
Table 3-5 SPS—8 Experiment Design Matrix .................................................................... 47
Table 3-6 Distribution of SPS-8 ﬂexible pavements sites by subgrade type and climatic
zone .................................................................................................................. 47

Table 3—7 Distribution of SPS-8 ﬂexible pavements sections by design, subgrade type-

climatic zone .................................................................................................... 48
Table 3-8 Details of instrumented sections for ﬂexible pavements .................................. 51
Table 3—9 Instrumentation details for all the SPS-l sections in Ohio ............................... 51
Table 4-1 Categorized list of variables for ﬂexible pavements (SPS-l and SPS-8) ......... 55
Table 4-2 Summary of SPS-l data elements availability ............................... ' .................. 6 1
Table 4-3 KESAL per year for SPS-l Experiment ........................................................... 64
Table 4-4 Intended SPS-l site factorial [1] ....................................................................... 67
Table 4-5 SPS-l site factorial —— From DataPave 3.0 ...................................................... 69
Table 4-6 Summary of comparison between target and as-constructed layer thickness .. 72
Table 4-7 Controlled vehicle parameters .......................................................................... 93
Table 4-8 Series 11 Truck Parameters — ODOT Single-Axle Dump Truck ...................... 93
Table 4-9 Series 11 Truck Parameters — ODOT Tandem-Axle Dump Truck ................... 93
Table 4-10 Series IV Truck Parameters - ODOT Single-Axle Dump Truck ................... 93
Table 4-11 Summary of SPS-8 data element availability -F1exible pavements ............... 96
Table 4-12 Summary of Environmental data of the sections in SPS-8 ............................. 98
Table 4-13 Subgrade soil properties for SPS-8 ﬂexible pavements ................................. 99
Table 4-14 Construction details of the ﬂexible pavement sections in SPS—8 ................. 103
Table 5-1 Calculated performance indicators ................................................................. 114
Table 5-2 Operational signiﬁcant differences for various performance measures ......... 130

Table 5-3 Actual replication of sections within SPS—l experiment— Fatigue cracking 139

Table 5-4 Calculation of standard deviate for alligator cracking - Alabama (1) ............ 143
Table 5-5 Site Level Comparisons for the SPS-l Experiment ....................................... 149
x

 

«1"1

Table 5-6 Example calculation of relative performance (State l-Alligator Cracking)... 150
Table 6-1 Identiﬁed sites and sections with rutting problems ........................................ 163
Table 6-2 Average asphalt mixture properties in the ﬁeld ....... A ...................................... 163

Table 6-3 Summary of p-values from AN OVA for determining the effect of main design
factors on pavement rutting ............................................................................ 167

Table 6-4 Summary of marginal means from ANOVA for determining the effect of main
design factors on pavement rutting ................................................................ 167

Table 6-5 Summary of p-values from AN OVA for determining the effect of experimental
factors on pavement rutting ............................................................................ 168

Table 6-6 Summary of marginal means from AN OVA for determining the effect of
experimental factors on pavement rutting ...................................................... 168

Table 6-7 Subjective ratings of drainage functioning at SPS-l test sections based on
video inspection results (source: [4]) ............................................................. 174

Table 6-8 Summary of p-values (non-parametric test) for Site Analysis - Level-A ...... 186
Table 6-9 Summary of p-values (non-parametric test) for Site Analysis - Level-B ....... 191

Table 6-10 Summary of p-values from LDA for determining the effect of experimental
factors on pavement performance measures ................................................. 206

Table 6-11 Summary of p-values from BLR for determining the effect of experimental
factors on pavement performance measures (Wet zones) ............................. 209

Table 6-12 Summary of p-values from BLR for determining the effect of experimental
factors on pavement performance measures (All zones) .............................. 209

Table 6-13 Summary of p-values from AN OVA for determining the effect of main design
factors on pavement performance measures—Overall ................................. 215

Table 6-14 Summary of marginal means from ANOVA for determining the effect of
main design factors on pavement performance measures—Overall ........... 215

Table 6-15 Summary of p-values from ANOVA for determining the effect of design
factors on ﬂexible pavement perforrnance—WF Zone ................................ 218

Table 6-16 Summary of marginal means from ANOVA for determining the effect of
main design factors on pavement performance measures— WF Zone ........ 218

Table 6-17 Summary of p-values from AN OVA for determining the effect of design
factors on ﬂexible pavement performance—WNF Zone .............................. 220

Table 6-18 Summary of marginal means from ANOVA for determining the effect of
main design factors on pavement performance measures— WNF Zone ..... 220

Table 6-19 Summary of p-values from AN OVA for determining the effect of site factors
on pavement performance measures (Main effects only) ............................. 223

Table 6-20 Summary of marginal means from ANOVA for determining the effect of site
factors on pavement performance measures (Main effects only) ................. 223

xi

 

I.

Table 6-21 Summary of p-values from AN OVA for determining the effect of site factors

on pavement performance measures (With interaction effects) ................... 224
Table 6-22 Summary of marginal means from AN OVA for determining the effect of site
factors on pavement performance measures (Interaction effects only) ........ 224
Table 6-23 Summary of p-values from one-way AN OVA for determining the effect of
site factors on pavement performance measures .......................................... 229
Table 6-24 Summary of marginal means from one-way AN OVA for determining the
effect of site factors on pavement performance measures ............................ 229
Table 6-25 Summary of p-values for comparisons of standard deviates— Fatigue
cracking ......................................................................................................... 235
Table 6-26 Summary of means of PI for fatigue cracking .............................................. 236
Table 6-27 Summary of p-values for comparisons of standard deviates— Structural
rutting ............................................................................................................ 240
Table 6-28 Summary of means of PI for structural rutting ............................................. 241
Table 6-29 Summary of p-values for comparisons of standard deviates— Change in IRI
..................................................................................................................... 245
Table 6-30 Summary of means of PI for change in TRI .................................................. 246
Table 6-31 Summary of p-values for comparisons of standard deviates— Transverse
cracking ......................................................................................................... 249
Table 6-32 Summary of means of PI for transverse cracking ........................................ 250
Table 6-33 Summary of p—values for comparisons of standard deviates— Longitudinal
cracking-WP ................................................................................................. 253
Table 6-34 Summary of means of PI for longitudinal cracking-WP .............................. 254
Table 6-35 Summary of p-values for comparisons of standard deviates— Longitudinal
cracking-NWP ............................................................................................. 257
Table 6-36 Summary of means of PI for longitudinal cracking-NWP ........................... 258

Table 6-37 Summary of p-values from ANOVA for determining the effect of design
factors on ﬂexible pavement response —-— Overall ....................................... 263

Table 6-38 Summary of p-values from ANOVA for determining the effect of site factors
on ﬂexible pavement response — Overall .................................................... 263

Table 6-39 Summary of correlations for deﬂections and DBPs with fatigue cracking .. 271

Table 6-40 Summary of correlations for deﬂections and DBPs with rut depth .............. 272

Table 6—41 Summary of correlations for deﬂections and DBPs with IRI ....................... 273

Table 6-42 ‘Simpliﬁed’ summary of effects of design and site factors for ﬂexible
pavements ..................................................................................................... 301

Table 7-1 Summary results for comparison between similar sections (SPS-8 vs. SPS-l)
..................................................................................................................... 314

xii

LIST OF FIGURES

Figure 3-1 Geographical location of SPS-l sites .............................................................. 35
Figure 4-1 Summary of dependent and independent variables ......................................... 54
Figure 4-2 Type of available data in LTPP database ........................................................ 57
Figure 4-3 Data Extraction Process Flow Chart ............................................................... 58
Figure 4-4 Scatter plot showing site distribution by climate ............................................ 68
Figure 4-5 Frequency plot for actual AC thickness .......................................................... 72
Figure 4—6 Scatter plot for actual AC thickness by site .................................................... 73
Figure 4-7 Frequency and scatter plots for actual base thickness ..................................... 74
Figure 4-8 Occurrence of fatigue cracking — SPS-l experiment .................................... 76
Figure 4-9 Extent of fatigue cracking— SPS-l experiment ............................................. 77
Figure 4-10 Age distribution of all cracking distresses —— SPS-l experiment ................. 77
Figure 4-11 Fatigue cracking by site —- SPS-l experiment ............................................. 78
Figure 4-12 Occurrence of LC-WP — SPS-l experiment ................................................ 80
Figure 4-13 Extent of LC-WP -— SPS-l experiment ..................................................... 81
Figure 4-14 LC-WP by site -- SPS-l experiment ............................................................ 81
Figure 4-15 Occurrence of LC-NWP — SPS-l experiment ............................................. 82
Figure 4—16 Extent of LC-NWP —- SPS-l experiment ..................................................... 83
Figure 4-17 LC-NWP by site —— SPS-l experiment ......................................................... 83
Figure 4-18 Occurrence of transverse cracking —- SPS-l experiment ............................. 84
Figure 4-19 Extent of transverse cracking —- SPS-l experiment ..................................... 85
Figure 4-20 Transverse cracking by site — SPS-l experiment ........................................ 85
Figure 4-21 Extent of rut depth —— SPS-l experiment ...................................................... 87
Figure 4-22 Rut depth by site —— SPS-l experiment ......................................................... 88
Figure 4-23 Age distribution of rut depth measurement — SPS—l experiment ................ 88
Figure 4-24 Extent of AIRI— SPS-l experiment ............................................................. 89
Figure 4-25 Extent of IRIo— SPS-l experiment .............................................................. 90
Figure 4-26 Roughness by site —- SPS-l experiment ....................................................... 91
Figure 4-27 Age distribution of roughness measurement — SPS-l experiment .............. 91
Figure 4-28 Age of the ﬂexible pavements in SPS-8 ..................................................... 104
Figure 4-29 Age distribution in the SPS-8 sites — ﬂexible pavements .......................... 104
Figure 4-30 Distribution of distresses in SPS-8 ﬂexible pavements sections ................ 105

xiii

Figure 4-31 Distribution of IRI and Rutting in SPS-8 ﬂexible pavements ..................... 105

Figure 5-1 Fatigue cracking with age— AL ( 1) ............................................................. 109
Figure 5-2 Longitudinal cracking-WP with age — AL (1) ............................................ 109
Figure 5-3 Transverse cracking with age —- IA (19) ..................................................... 110
Figure 5-4 Rutting with age —— IA (19) .......................................................................... 110
Figure 5-5 Poor Performance.......' ................................................................................... 111
Figure 5-6 Good Performance ........................................................................................ 111
Figure 5-7 Comparing different performance curves — An Example ........................... 114
Figure 5-8 Summary of the model identiﬁcation process [5] ......................................... 116
Figure 5-9 Summary of research tasks ............................................................................ 117
Figure 5-10 Methodology for overall analysis ................................................................ 118
Figure 5-11 AN OVA diagnostics-Residual plot without transformation ....................... 123
Figure 5-12 ANOVA diagnostics-Residual frequency without transformation ............. 123
Figure 5-13 ANOVA diagnostics-Residual normality without transformation .............. 124
Figure 5-14 ANOVA diagnostics-Residual plot with log (natural) transformation ....... 124
Figure 5-15 ANOVA diagnostics-Residual frequency with log (natural) transformation
........................................................................................................................ 125
Figure 5-16 AN OVA diagnostics-Residual normality with log (natural) transformation
........................................................................................................................ 125
Figure 5-17 Performance criteria for fatigue cracking [l] .............................................. 127
Figure 5-18 Performance criteria for PI of fatigue cracking ........................................... 127
Figure 5-19 Performance criteria for rut depth [1] ......................................................... 127
Figure 5-20 Performance criteria for PI of rut depth ...................................................... 127
Figure 5-21 Performance criteria for transverse cracking [1] ......................................... 128
Figure 5-22 Performance criteria for PI of transverse cracking ..................................... 128
Figure 5-23 Performance criteria for roughness-Flexible [l] ......................................... 128
Figure 5—24 Type I and Type II Errors ............................................................................ 130
Figure 5-25 Effect of replication and variance on statistical power—One-way ANOVA
........................................................................................................................ 135
Figure 5-26 Effect of replication and variance on statistical power—Multi-Factorial
ANOVA ......................................................................................................... 138
Figure 527 Effect of mean difference and variance on statistical power—Test means for
two populations ............................................................................................ 140
Figure 6-1 Rutting with time for SPS-l pavements - All sections ................................. 164

xiv

Figure 6-2 Rutting with time for SPS-l pavements - Selected sections ........................ 164

Figure 6-3 Transverse proﬁle for base rutting—Section 20-0102 .................................. 165
Figure 6-4 Transverse proﬁle for asphalt rutting— Section 31-0113 ............................. 165
Figure 6-5 Transverse proﬁle for (HMA + base) rutting— Section 39-0101 ................. 165
Figure 6-6 Transverse proﬁle for (Base) rutting— Section 51-0113 ............................. 165
Figure 6-7 Rutting growth with time for SPS-l pavements — All sections .................... 166
Figure 6-8 Rutting grth with time for SPS-l pavements — Selected sections ............ 166
Figure 6-9 Fatigue cracking with time for SPS-l pavements — All sections .................. 170
Figure 6-10 Fatigue cracking with time for SPS-l pavements - Selected sections ........ 170
Figure 6-11 IRI with time for SPS-l pavements - All sections ..................................... 171
Figure 6-12 Transverse cracking with time for SPS-l pavements — All sections .......... 171

Figure 6-13 Longitudinal cracking-WP with time for SPS—l pavements — All sections 172
Figure 6-14 Longitudinal cracking-NWP with time for SPS-l pavements — All sections

..................................................................................................................... 172
Figure 6-15 Methodology for site level analysis (SPS-l) ............................................... 185
Figure 6-16 Effect of experimental factors on fatigue cracking ..................................... 199
Figure 6-17 Effect of experimental factors on rutting .................................................... 200
Figure 6-18 Effect of experimental factors on roughness ............................................... 201
Figure 6-19 Effect of experimental factors on transverse cracking ................................ 202
Figure 6-20 Effect of site factors on longitudinal cracking-WP .................................. ... 203
Figure 6-21 Effect of site factors on longitudinal cracking-NWP .................................. 204
Figure 6-22 Effect of design factors on fatigue cracking ............................................... 234
Figure 6-23 Effect of design factors on structural rutting ............................................... 239
Figure 6-24 Effect of design factors on change in IRI .................................................... 244
Figure 6-25 Effect of design factors on transverse cracking .......................................... 248
Figure 6-26 Effect of design factors on longitudinal cracking-WP ................................ 252
Figure 6-27 Effect of design factors on longitudinal cracking-NWP ............................. 256
Figure 6-28 Fatigue cracking and SCI relationship— State (20) Kansas ........................ 269
Figure 6-29 Fatigue cracking and AF relationship— State (20) Kansas ........................ 269
Figure 6—30 Roughness and BDI relationship— State (20) Kansas ................................ 269
Figure 6-31 Rut depth and BDI relationship— State (20) Kansas ................................ 269
Figure 6-32 Roughness and BDI relationship— State (39) Ohio .................................. 269
Figure 6-33 Rut depth and AF relationship— State (39) Ohio ....................................... 269

XV

 

Figure 6-34 Relationships between age and different performance measures ............... 276
Figure 6-35 Apparent relationships between SCI and fatigue cracking ......................... 277
Figure 6-36 Apparent relationships between AREA and fatigue cracking ..................... 278
Figure 6-37 Apparent relationships between AREA and longitudinal cracking-WP ..... 279
Figure 6-38 Apparent relationships between AREA and longitudinal cracking-NWP .. 280

Figure 6-39 Apparent relationships between AREA and transverse cracking ............... 281
Figure 6-40 Apparent relationships between BDI and rut depth .................................... 282
Figure 6-41 Apparent relationships between BDI and IRI ............................................. 283
Figure 6-42 Relationship between measured responses and observed performances—
Fatigue cracking and rutting ........................................................................ 286
Figure 6 43 Relationship between measured responses and observed performances—
Roughness .................................................................................................... 287
Figure 7-1 Average fatigue cracking by experimental factors— SPS-8 ﬂexible pavements
..................................................................................................................... 310

Figure 7-2 Average fatigue cracking by subgrade type— SPS-8 ﬂexible pavements.... 310
Figure 7-3 Average LC-WP by experimental factors— SPS-8 ﬂexible pavements ....... 310

Figure 7-4 Average LC-WP by subgrade type— SPS-8 ﬂexible pavements ................. 310
Figure 7 -5 Average LC-NWP by experimental factors— SPS-8 ﬂexible pavements... 311
Figure 7-6 Average LC~NWP by subgrade type— SPS-8 ﬂexible pavements .............. 311

Figure 7-7 Average transverse cracking by experimental factors— SPS-8 ﬂexible
pavements ....................................................................................................... 31 1

Figure 7-8 Average transverse cracking by subgrade type— SPS-8 ﬂexible pavements311
Figure 7-9 Average roughness by experimental factors— SPS-8 ﬂexible pavements... 312
Figure 7-10 Average roughness by subgrade type— SPS-8 ﬂexible pavements ........... 312
Figure 7-11 Average rut depth by experimental factors— SPS-8 ﬂexible pavements... 312

Figure 7-12 Average rut depth by subgrade type— SPS-8 ﬂexible pavements ............. 312

xvi

CHAPTER 1 - INTRODUCTION

1.1 INTRODUCTION

Considerable progress has been made over the past 50 years in the ﬁeld of
pavement engineering towards providing economical pavements. Pavement design
and analysis methods, construction practices, and material speciﬁcations have been
developed and reﬁned as tools to facilitate engineers in the design and management of
highway infrastructure. While much has been learned in recent years, still there is a
demand from highway agencies to improve existing procedures and tools for effective
management of pavement networks. These needs from the highway agencies to
explore more innovative design procedures will help them in meeting the ever

increasing highway infrastructure demands within limited resources.
1.2 PROBLEM STATEMENT

At present, highway agencies lack adequate information on the inﬂuence of
pavement drainage on the performance of ﬂexible pavements. There is very limited
experimental ﬁeld data to quantify its inﬂuence or effect on pavement performance.
Currently, the inﬂuence of drainage is incorporated into the AASHTO Design Guide
through a modiﬁed structural number concept using subjective judgment for the
selection of the layer modiﬁed coefﬁcients. As stated in Volume II, Appendix DD of
the AASHTO Design Guide (1986), these coefficients were developed through a
theoretical approach because very limited ﬁeld data existed at that time [1].

The modeling of the pavement structure as an elastic layer system has been the
main focus of the research in the past two decades. The pavement layers are
characterized by their thickness, stiffness and Poisson’s ratio with stiffness (modulus)

being the most important parameter. The information on interactions of these key

factors with other variables such as climate is not well understood and therefore, still
needs to be explored in order to improve relationships between pavement
performance and response. Therefore, controlled ﬁeld experiments are necessary to
answer the following questions [1]:

0 “To what extent does the influence of pavement design on pavement
performance vary from wet to dry climatic zone and what is the
relative importance in each zone?”

0 “To what extent does asphalt concrete thickness, type or thickness of
the base layer inﬂuence pavement performance and what is the relative
importance of each? How do environmental factors inﬂuence the

relative importance?”

To address the above questions, the Long Term Pavement Performance
(LTPP) Speciﬁc Pavement Studies (SPS) 1 and 8 experiments were designed to
provide information on the relative merits of different design features in newly
constructed ﬂexible pavements for achieving different levels of performance under
heavy trafﬁc (SPS-l) as well as in the absence of heavy trafﬁc (SPS-8). Typical
features include hot mix asphalt (HMA) surface layer thickness, base type and
thickness, and drainage (presence or absence thereof). In addition to this,
instrumented sections were included in the SPS monitoring sites located in Ohio
(SPS-l).

For speciﬁc site conditions (e.g., trafﬁc level, climatic conditions, and
subgrade type), the response and performance of ﬂexible pavements will depend not
only on pavement layer thicknesses and material properties, but also on other design

and construction features (e.g., presence of in-pavement drainage and base type, etc.).

Recent research based on limited analyses has documented the effects of these
features on pavement response (as measured by deﬂection and strain) and the
contributions of these features to achieving different levels of performance (as
measured by type and extent of distress or smoothness).

The importance of this experiment is highlighted by its ability to evaluate the
interaction of drainage, structural parameters, subgrade type and climatic factors on
pavement performance in a controlled manner. The data available from the LTPP
studies, including instrumented SPS-l test sections in Ohio are expected to provide
the information needed for a more rigorous analysis to enhance understanding of the
effects of these features on ﬂexible pavement response and performance and to
deve10p well-supported conclusions regarding their inﬂuence.

There is therefore a need to determine the effects of design and construction
features on ﬂexible pavement performance and response, and to establish their
relative importance. This research should provide guidance for identifying appropriate
features for different pavement types, preliminary information on the relationship
between pavement response and performance, and recommendations for improving

data collection activities.
1.3 RESERACH OBJECTIVES

The purpose of this research is to more precisely determine the relative
inﬂuence of speciﬁc factors that affect the performance of ﬂexible pavements. The
factors addressed in this study include drainage, base type and thickness, and asphalt
surface thickness. The study goals also include a determination of the inﬂuence of
environmental region and soil type on these factors. Accomplishing these objectives
will provide substantial improved tools for use in the design and construction of new

and reconstructed ﬂexible pavements. The speciﬁc objectives for this research are:

1) To determine, for speciﬁc site conditions (climate and subgrade type)
the contributions of design and construction features to achieving
different levels of performance.

2) To determine, for speciﬁc site conditions, the effects of design and
construction features on pavement response (deﬂections and deﬂection
basin parameters), and

3) To provide information on the apparent relationship between pavement

response and performance.

The research is limited to new (i.e., non-rehabilitated) ﬂexible pavements. The
research is based on the data available from the LTPP experiments SPS-l (strategic
study of structural factors for ﬂexible pavements), and SPS-8 (a study of
environmental factors in absence of heavy loads). The analysis is limited to using the
data available in the LTPP Information ManagementSystem (IMS) database
classiﬁed as "Level E" (Release 17) and response data available from LTPP

instrumented test sections.
1.4 SCOPE OF STUDY

The scope of the study included the review and analysis of LTPP data
(DataPave) pertaining to the SPS-land SPS-8 experiments. All relevant data for these
experiments were obtained and reviewed from Release 17.0 (January 2004) of the
IMS database. After the data were obtained, a relational database was prepared for
analyses of the study. Based on the availability of the data and the extent (and
occurrence) of distresses, apprOpriate analyses (site-level and overall analyses) were
conducted to fulﬁll the objectives of the study. At the site-level analysis each site is

considered separately and the consistency of the effects across the sites is studied. The

4

overall analyses are conducted, using the wealth of data from all the experiment
sections in order to draw broad conclusions. Attempts are also made to verify
apparent relationships between response (Falling Weight Deﬂectometer data and
Dynamic Load Response data) and performance of the test sections in the SPS-l
experiment. Based on all analyses (site-level and overall), the effects of design and
construction features on pavement performance and response are studied. Finally,

recommendations for future research and data collection are given.
1.5 REPORT ORGANIZATION

The report is divided into eight chapters including this introductory chapter.
Chapter 2 highlights the summary of ﬁndings from the LTPP related studies, which
can be useful for this research. The detailed descriptions of the experiment designs
and the current status of SPS-l and SPS-8 experiments are presented in Chapter 3. A
summary of data availability and extent and occurrence of distresses is presented for
the two experiments in Chapter 4. Based on the extent and occurrence of distresses,
different methods of analysis were employed. A brief description of each of these
analytical methods and methodology used for the research are detailed in Chapter 5.
Chapters 6 and 7 are summaries of results from all analyses conducted on SPS-l and
SPS-8 data, respectively. A synopsis of the salient ﬁndings from all the analyses, for
each experiment, is presented in Chapter 8. Finally, based on the experience with
LTPP data gained from this research, recommendations for future data collection and

research are presented.

CHAPTER 2 - LITERATURE REVIW

2.1 GENERAL

Pavements are designed to fail after sustaining the anticipated design trafﬁc loads.
The damage due to the repetitions of heavy axle loads accumulates over time and
eventually the pavement structure needs to be rehabilitated after a certain threshold for
performance is exceeded. The performance of a particular pavement section depends on
the interaction of many factors including structural design parameters, material
properties, the operating environment and most importantly the amount and magnitude of
trafﬁc loads. Thus a pavement structure should be considered as a system, where

different components play their individual roles.

Traditionally, the performance of various pavement materials can be investigated
through the use of small—scale laboratory tests, observing ﬁeld performance or both.
Small-scale laboratory investigations may help in measuring basic material properties and
understand the material behavior under controlled conditions; however, this might not be
sufﬁcient to capture the actual in-situ material behavior because of different boundary
conditions in the ﬁeld and the interaction of many factors at the same time. Therefore,
researchers have also used full-scale accelerated testing to simulate the ﬁeld boundary
conditions and estimate the long term performance of pavements in a short period of
time. However, the results of accelerated full-scale tests areialso dependent on how close
this simulation is to the actual ﬁeld conditions in terms of materials, environment and
loading conditions. On the other hand observing ﬁeld test results from in-service

pavements provide a good indication of the short and long term performance potential

[1]. This approach is more realistic, but demands long term monitoring and may become

very expensive.

Many ﬁeld tests [2] have been conducted in the past to monitor long term
pavement performance to help develop a pavement design procedure. For example the
design procedure recommended by the American Association of State Highway and
Transportation Ofﬁcials (AASHTO) is based upon the results of the extensive AASHO
Road Test conducted in Ottawa, Illinois, in the late 1950’s and early 1960’s [3].
However, most of these ﬁeld tests were conducted several decades ago and had various
limitations such as short-term monitoring of pavements (1 to 2 years), limited climatic
conditions, limited materials and most importantly limited truck loads and axle
conﬁgurations. In order to address all these limitations the US Congress approved the
Strategic Highway Research Program (SHRP) in 19.87 [2]. This included the Long Term

Pavement Performance (LTPP) program, described next.

2.2 BACKGROUND

Understanding "why" some pavements perform better than others is key to
building and maintaining a cost-effective highway system. In 1987, the Long-Term
Pavement Performance (LTPP) program —— a comprehensive 20-year study of in-service
pavements — began a series of rigorous long-term ﬁeld experiments monitoring more
than 2,400 asphalt and portland cement concrete pavement test sections across the US.
and Canada [4]. Established as part of the Strategic Highway Research Program (SHRP)
and now managed by the Federal Highway Administration (FHWA), LTPP was designed

as a partnership between FHWA and the States and Provinces. LTPP's goal is to help the

States and Provinces make decisions that will lead to better performing and more cost-

effective pavements.

In the LTPP program various experiment designs were planned to conduct
research on a variety of factors affecting pavement performance. The experiment designs
were divided into two main categories: general pavement studies (GPS) — focusing on
the most commonly used pavement designs (existing pavements), and speciﬁc pavement
studies (SPS) — to study certain pavement engineering factors (new pavements). Within
these two categories, further experiment designs were planned to study different types of
pavements, etc. Two of the experiments in the SPS categories for ﬂexible pavements will
be the focus of this research. These are the SPS-l and SPS-8 experiments, which are
described in Chapter 3. In this Chapter, the previous research conducted using LTPP data,

' especially SPS-l and SPS-8 experiments is brieﬂy presented.

2.3 SUMMARY OF RELATED LTPP STUDIES

This section summarizes the ﬁndings from the literature review of research
reports that deal with pavement performance in the ﬁeld. The review included
FHWA/LTPP reports, NCHRP reports as well as additional literature, and was focused
on research that has identiﬁed factors affecting pavement response and performance
including roughness. The most relevant reports were found to be those from studies

addressing the Long Term Pavement Performance (LTPP) experiments.

The information obtained from the literature review has been used to identify
various factors that have been shown in past research as having an effect on response and

performance progression. The brief ﬁndings from these studies are presented next.

2.3.1 Factors Affecting Flexible Pavement Performance
A study [5] entitled “Structural Factors for Flexible Pavements” was conducted
using SPS-l data. The summary of ﬁndings from this preliminary study is given below:
Layer Thickness
- The SPS-l test sections with thick (1 78-mm (7-inch)) AC surface layers appear to
be smoother and develop less fatigue cracking than those sections with thin (102-

mm (4-inch)) surface layers. This conﬁrms a similar ﬁnding from earlier studies.

0 In the SPS-l experiment, AC surface thickness and the age of the project appear
toinﬂuence the amount of fatigue cracking that occurs. The test sections that are
younger and have thicker AC surface layers have the least fatigue cracking.
Base'Layer

- Hot-mix asphalt (HMA) pavements with unbound aggregate base layers show
greater rut depths than those sections with asphalt-treated base layers. This
suggests that a portion of the rutting measuredat the surface is a result of
permanent deformations in the unbound aggregate base layer, which is consistent

with a previous ﬁnding from analysis of the GPS test sections.

0 The HMA pavements with unbound aggregate layers have slightly more fatigue
cracking and higher IRI values than those sections with asphalt-treated base

layers.

0 The test sections with coarse-grained soils, asphalt-treated base layers, permeable

base layers, thicker bases, and thicker HMA layers were found to be smoother.

o The test sections with permeable asphalt-treated base layers exhibit more fatigue

cracking than those without permeable base layers.
9

Subgrade

0 HMA pavements built over coarse-grained subgrade soils are smoother than
pavements built over ﬁne-grained subgrade soils. This is consistent with the
ﬁnding in the SPS-2 JPCP: A stiffer foundation contributes to smoother

pavements.

o HMA pavements built over coarse-grained subgrade soils and in a no-freeze
climate are smoother and stay smoother over a longer period of time than do those
built over ﬁne-grained subgrade soils in a freeze climate. HMA pavements built
over ﬁne-grained sub-grades and in a wet-freeze climate are substantially rougher

than those built in other climates.

o HMA pavements built over ﬁne-grained subgrade soils have more fatigue

cracking than those projects built over coarse-grained subgrade soils.

0 Subgrade soil type and, to a lesser degree, age are important to the amount of
transverse cracking measured at eachsite. More tranSVerse cracking has occurred
on the HMA pavements built on ﬁne-grained soils than on pavements built on

coarse-grained soils.

Two studies[6, 7] were conducted to investigate the effects of sub-drainage on the
performance of asphalt and concrete pavements. The following is a summary of their

ﬁndings for asphalt pavements:

0 Based on 7 years (on average) of ,SPS-l data, those HMA sections built on
permeable bases without edge drains were found to perform better than those with

edge drains.

10

o The ranking of performance in terms of IRI and cracking for various base types
with all other design features matched is from poor to good performance: un-
drained dense-graded aggregate bases, drained permeable asphalt-treated bases,

and un-drained dense-graded asphalt-treated bases.

0 The results in terms of rutting for the above three sub-drainage designs were

inconclusive.

A fourth study was conducted on the environmental effects in the absence of
heavy loads using SPS-8 data [8], however, the ﬁnal report of this study is not published.
The SPS-8 experiment can be considered as an extension of SPS-l (new ﬂexible
pavements) and SPS-2 (new rigid pavements) with limited trafﬁc effects. The preliminary
conclusions from this researCh are summarized below:

Temperature and Precipitation

0 For SPS-8, asphalt concrete (AC) sections, the most prevalent early distress is
longitudinal cracking outside the wheel path. The distress is most commonly
observed for sections built in the wet-freeze climates and for sections on an active
subgrade (frost-susceptible or swelling soils due to freeze-thaw cycles). Fatigue,
longitudinal cracking in the wheel path, and transverse cracking are present on

just a few sections. The mean rut depths for all AC sections are below 6

millimeters (mm) (0.24 inches).

Subgrade

o Pavements (ﬂexible or rigid) constructed on active subgrade have the highest
mean initial International Roughness Index (IRI) values and slopes (the

smoothness rate of change over time), followed by pavements constructed on ﬁne

11

subgrade and then pavements constructed on coarse sub-grade. The data support a
similar ﬁnding from previous studies that a good working platform (speciﬁcally,
stabilized base and granular subgrade or embankment) contributed to a smoother

pavement construction.

Initial IRI values for the SPS-8 test sections show that ﬂexible pavements were
constructed to be smoother than the rigid pavements. The analysis of IRI slopes
indicates that the subgrade is the most important factor for ﬂexible sections, while

precipitation appears to be the most important factor for rigid sections.

Pavement Type and Layer Thickness
The SPS-8 ﬂexible pavements with thin (102-mm (4-inch)) AC surface layers

were found to be smoother than the sections with thick (178-mm (7-inch)) AC

layers.

Comparisons of SPS-I, -2, and -8 Test Sections 2
As expected, trafﬁc loading is much heavier on SPS-l and SPS-2 than on SPS-8

sites. As of June 2001, the estimated accumulated ESALs on SPS—l sites was

about 1.46 million, compared to 0.043 million ESALs on SPS-8 AC sites.

The average IRI slopes (the smoothness rate of change over time) for both SPS-l
and SPS-2 sections are much higher than for the corresponding SPS-8 sections.

The variability of mean IRI slopes is higher for PCC than for AC sections.

Overall, the much more heavily loaded SPS-l and SPS-2 sections exhibit higher
amounts of load-related distresses. Such distresses include AC rutting, AC fatigue

cracking, JPCP joint faulting, and JPCP transverse cracking. However, the non-

12

load-related distresses including AC transverse cracking and non-wheel path

longitudinal cracking are similar for SPS-l, SPS-2, and SPS-8.

Another study [9]was conducted using SPS-l data to identify the factors affecting
pavement smoothness. A summary of the main ﬁndings from this research is given
below:

0 A signiﬁcant difference between early age IRI of pavements placed on DGAB
and ATB was observed. No signiﬁcant difference in early age IRI was obtained
on pavements placed on PATB when compared to the other two base types.

0 The SPS-l projects that showed the highest increase in IRI were located in
Kansas, Iowa and Ohio, reasons being fatigue and transverse cracking for KS
(20), transverse and longitudinal cracking in the wheel path (WP) for IA (1.9), and
rutting for OH (39) site. Some of the test sections in TX (48) are showing higher
increase in IRI of over 10% within an approximate 6-month period, which is
attributed to rutting. _

0 Although the pavements in IA (19), KS (20), and OH (39) achieved a smooth
pavement initially, many sections, including very thick sections had high
increases in roughness during the initial life of the pavement. Achieving a smooth
pavement initially does not guarantee that it will remain smooth even during the
initial life.

0 Mix design problems in the AC, inadequate preparation of the subgrade prior to
placing the pavement, or other construction problems can cause smoothly built

pavements to have higher increase in roughness within a short time period.

Table 2-1 summarizes the effects of the structural and site factors on ﬂexible

pavement performance from other relevant research conducted using the LTPP data.

13

 

:5.

.0>0.m com... &:.m5q05&5m .0 055005305 3.5.1.

w5n5:< 0. 005.0. 3&55 53 m..0m :0:.w& 0:... 5>0 50:08 . mmO :. :.30..& .y:
.&:.0.005 2&5. 0.0:. :. .55.. 5:.0> 0&85050 5.5.5500 055.05 530...
5559.0: .0. 5.50.00 05 0505:. 0. 00.0005 0n. 5:: 5.. .&:.0.55

0053.5... .05 &:.0.005 2&5. .&:.§: .0. 50.0 .0: 05 5.5.wbsm 305.0 .0 000.5 0:...

5.5.005
05.31%

 

:5.

.:03 5 :0.5..%00:m30..&... 0. 5.00.05. 053 5:. 50.50... 5.. 00506:. 53 00555.5:
55: .5530: .02.; E. 55.: 5: b.055m m.:0E:0..>:0 .0: b0> :. m:0..00m
.52.; 3.. 55.: .5: 5.05 355.058. .0 505.:

:w... a .0 50:. 050:. :0sz m .5: 5... 505 :. 35:55: 55:50:35 .0550... :.

. 50.22.3005

55.: .0 505 :. 050.. 053 &:.0.055 0&5. .05 50:520. 55.: 5.5550

. 55:55..

5.2.5: .05 0.5.0.005 02055.. 0&5. .&:...E 0.0:. 005.595 0. .05. 55.5.0 53 mocoN 0.88:0

. 55:550. .0 05. 530&

55.: m 05: .05 5.0.0.5 8535:. 0.0:. 005.595 555:0 5200 .5555 :53
:. 00.0005 50. 05 5.0.55 .5 02:3 .&:.5. .05.. 0. 0050005 05 5:55:50. :w... 0.5
.Uomm 0>05 m5: .0 53:5: :5 055595. 50:. .555 5.0.8 3.2.550... .0 5:55:
A050: m..m:0> 050..-:05 0.5 .0 :0500. .3 50.5.0055 m. n7:1. :. 055.5050.
03.3 60.0 .0 53. 0.5 .0 :0500. :5 m5: 53 .0 505:: 0... .3 05.5.0055

m. 00.: :. 055.02 5555050. :5 055.05 0.5.0:. 5.5.03 0.55:0 .05.).

50.5500

 

.o. .0.

.&:.0.0:..0 5.50.5:

:5 0.0.555: 00505:. 0. 05:..500 5:. 0.2mm. 00505:. 55:03:00.: 20.0....
5:005:00 .05 :95: 0. 0.6 3.0.0.05 2&5. 0.0:. b.50....:&.m 05: 50 mu.<mm.

530. :0 0:5.0. 0:. 5.3 .5505: 5>.& 0 5... 0m 50550.5: 2&5. :. 5055...:

&.n 0:. 5.. 03.5005. 0.. 50 5.5.23 550 5:05.05 5:. 555 .. :5 .&:.0.05.0 2&5. @5504 SEE...
05.50 .0: 0.0 3.4mm. 5:. 505 .0: 50: m2. .m5_0:.5>0Z 25.550 U< 5.5.0.5:
3.000 .05 .000& 503.0: 505...: 3.50.0.5... 0: 0. .050. .0: 053 mq<mm 0355500
.5530: .90... 50:500. 9... :5 &:.0.55 2&5. A... .&:..5.. 3 &:.505:. :.
.50....:&.m 0: 0. .050. 53 A3<mm 035.2550. 50.2.05. 5.5 .95: .0 50:5: 55....

 

 

mDUGOHDm 0M

 

05053. .05 0055.85: .5805“. :0 .00....m 0.5.55 .055”.

 

 

 

 

35:55: 03.05.“. .0 00555.55. :5 05005”. 0... &:..00...< £0.05... .0 35855 ..N 0.5%

14

 

 

 

555.... 50.00 0... c. 35:00: .000... $.50 .0

 

 

 

 

 

 

 

 

 

 

 

no 2 .55. 0:. 0: 3:... 2:. .5530... .&:...0. .0. .50...:w5 0: 0. 0:00. .0: 53 3.5085 3...m00m.>
55.0.5.0 05:. 0505:. :.3 00:05 3.5005? ..:..0m<
.:w.: .0 0m: 5:. 0.5.0.005 05555.. 0:: 00&..:. .0 5.005 .00.. m:0..:0.0:. 0.: 05...: 00.5000... .:..0.:.>.
mo: 5.0. 50000 :. .55. 5:0 0...:0 3: 0000:0500 .0:: m0.0> .5 30. 0...).
00... 0000:0500 500.0 0:: :m500 0.0.0:... :w00.:. 00:00:00 0: 505 m0.0> ..< 0< 0. m0.o> ..<
£5.55. ..0:. :. 0.: 0:. .::. E...:00 0. 00.00 5:.
mm 2 0: 0.00:: 5.0.00. 0&::.:.0 0>..500 :..3 5.5 .0.0.05:... 3.5020 &:.:0..0:0. 5. 0m::.:.
E0535 0&::.:.0 03.500 0:. 5... 00:00.55 0:. 0. 005.0. 0: 3:E 5...... 5.55.505 . Q
N 0:: . w0o :. 02:05.50: .0: 0.03 55.00. 0&::.:.0 02.500 .0 50.05: 0: ... 505:0...
. 05.. 5.... .0.
555.05 &:..00:.&:0 0: m. 05:... 555.5.0 00.0: 0:: 555.000. 0505:. 0. 050. 02.03.5000
8 2 5:3 0000:0500 00505:. 500.505 m0O .0. .5530... 0000.00... .0: 53 0&::.:.0 0050:9000
50.05:: m: &:0. 5 60555.50 .0. 000& 3..:.0:0& 53 .:..0.:E 05:: .:.0::.&
.0. :0..0:0E00 &:5:0.0:. 5:. 005.505 5:0 E0... 005:0: :00: 3..:5:0& 5:: ..
S E .mZm .0:&.: 0:: 5:. 500.500 :..3 0050500 5:3 Ava 505.05...
.y: .0 0505:. .0 2:. .0:&.: 0:.. 52m .030. 0:: .0:. 000.50: 0.00.8.0 . m0m 0. . .:0E0>:n.
00555.5 5.5 .5 ..0:.
8 L .0. 0.:5505. :00: 03.: 3:E 5:0..005 N EU 0:. :. 05:: 00.5.. .0 00:05.5 0:... 505. 0. .. H. 0 m: m
”N E 50.05 0.: 05.5.0 .0:.0 :0 m.00..m. 05:00.00: 0.: 50:55.0 0:: 50500.0 . 005.00.”. :w50Q
5.55:. 0:. m: &:0_ m: .&:...0. :. 050.000 : 0. 0:0. 3:E 555.02. 05: 0:. &:5:0.0:. .
050.50: 0.: 00.50.5000 0::
mo 2 0&500 0.00:5 .0:. 5.50.5: w...0.0:.0 00555.. &:5:0500 :. 0.0: 3.:::0.0 0.003 505.03...
05 555.050. 0:: &:.0.0:.0 00&..:. .&:....0. 050.000 0. 00.0005 0: 3:E 55:...5 ..:..Qm<
5.0.00.5 .:0E0>:0 ..:.0>0 0... “55:05:. .0 053:. 0.05000 ..::05 50.0.... w:5:
3 Z .&:.0.0:.0 05555.. .0. 5.0:. .::0...:&5 .5... 0:. 0: 0. 0:00. 53 0.0:... 0w< 5.05.0000 8%
50:05.00 05:005.: 0:: 00555.50 .:0E0>:0 :0 .00.... 0.::..:> 5.0:.

 

 

 

 

 

A 000:..50. 35:55:. 2:05.... .0 00585.5... 05. 052.5% 0:. &:..00..< 5.0.0:... .0 355:.0m ..m 0.5...

15

 

 

 

Ho L .3338 235 Sn 258 225 $.32 3:85 .8me 095 8mm
5.25% 253:: so 3.2: =9: 23% a: .532 BE 83 933.: 22:8 53, mEoEoSm
Ho 3 dog—Eaton Stop a no; 28:8 8:588 023 .532 5.3 33:00

mcozuom ESE: 2 3:22.85 on 9 258 $3 33 52:8» 05 mo E850 2829: 2:

DEAN—£08 omwm

 

.mcosoom mac 5 Enocawa

 

 

 

 

 

Ho 3 up 8 958 8: was 35:88 95:: 2: :o 3:69: 25:88 3.32% 2: no 80km mince—2 ommm mvmtoaohm.
.8me .5955. :5 35:5 5:: mac—88 0:95 82: £25 53 393 x83 x02... Riga—2

.wExoﬁo oawce .8 958 225 8:52 3:85 .3169:

no ~ H 22:8. 3:E; ms :25 $8on .550 no 8555 2: 9 26 2. EEE 855.8th coow cosmcmuw

$5 .5330: .oocactotoa wast: .52— 5; 805 :2: EEde can EEEd .EEnmvd vamwawwdx
.EEmné 5::de 353 9.3me .x. 29: can chE E 355.8th uoow 5:: 85838
.wcioﬂu 2.33 32 3m:

8 Z 8 22 3:62: “22:8. 85$: 53 mcocuom .335qu maﬁa 2: magic. 35 can :98. m2:on U<
83 5::an 05 E v.58: 832 E :32 2:95 .98— U< 2: .«o 3:62: “5:62 “0:3:

moocouomom omcoamom new 028::ng E0535 so Scam 28x33 56“,»

 

 

 

 

A 2625353 358025 0::on mo oozagotom Ea omcommum 05 mascot/x £8on .8 $2.83 Tm Bomb

16

IN"

2.3.2 Relationship between Pavement Response and Performance
In the NCHRP Report No. 10-48, “Assessing Pavement Layer Condition using
Deﬂection Data”[14], the performance assessment of the pavement layers using

deﬂection data has been investigated. Some results of this study are enumerated below: “

o For full depth and aggregate base pavements, Eacl can be used as an indicator to
detect cracking and stripping in the AC layer. EDI2 and Eabc3 were found to be
good condition indicators for base layer condition, while BCI4, eggs, SSR6 and
Egg7 appeared to be good condition indicators for the subgrade. For intact
pavements, the pavement overall fatigue cracking and rutting potentials are
mainly controlled by sac, Eabc and 85g.

0 Temperature adjustment is an important procedure in assessing the condition of
asphalt-surfaced pavements. Temperature adjusted indicators were found to be

able to predict fatigue cracking and rutting potentials of full depth and aggregate

base pavements.

- For full depth and aggregate base pavements, the analysis from both synthetic
data and ﬁeld data showed that SCI8 can be used to predict the AC modulus.

Also, high correlation was found between EDI and sac.

. The deﬂection basin parameter BDI can be used to assess upper layer condition in

cement treated base (CTB) pavements.

 

I AC modulus

2 BDI=D12-D24 (Base Damage Index)
3 Compressive strain at top of base layer

" BCI=D24-D36 (Base Curvature Index)
Compressive strain at top of subgrade

6 SSR=od/qu (Stress Parameter for rutting potential of subgrade)

Resilient modulus for subgrade
8 SCI =Do_-D.2 (Surface Curvature Index)

17

o Deﬂection values from the sensor four feet from the FWD load center (D48) can

be used to estimate subgrade condition in CTB pavements.

A recent study [15] on evaluating the feasibility of using FWD deﬂection data to
characterize the pavement construction quality using SPS-l data concluded that “FWD
can be effectively used to assist in the characterization of pavement quality for new or
reconstructed pavements during construction”. Furthermore, “quality” can be directly

’ equated to stiffness or modulus, and to other deﬂection-based parameters. It is therefore

expected that these parameters should correlate to the future performance of pavements.

Equation (2-1) can be used to calculate the area under the deﬂection basin using

the ﬁrst four sensors equally spaced at 12 inches from each other [15, 16].

Area=6* 1+2 5&- +2 9,—2— + 5 ' ‘ (2-1)
do do do
Where;
Area = The “Area " beneath the ﬁrst three 3 feet (914 mm) of the deﬂection basin;
d0 = FWD deﬂection measured at the center of load plate;
d1 = FWD deﬂection measured at 12 inches from the center of plate,-
d 2 = FWD deﬂection measured at 24 inches from the center of plate;
d3 = FWD deﬂection measured at 36 inches from the center of plate.

Equation (2-1) has been used for a load plate diameter of 12 inches. A maximum
Value of 36 is possible for “Area” when all the deﬂections are the same in Equation (2-1).
This will represent a very stiff pavement (extreme value). It was found using the forward
analysis[15] that when all the layers in a ﬂexible pavement were assumed to have the
same stiffness/modulus, the “Area” term is always equal to 11.037. Therefore, it was
concluded that if FWD deﬂection measurements carried out in the ﬁeld result in an

“Area” value close to 11.037, then the effective modulus or stiffness of the upper layer is

18

identical to that of the underlying layer(s). Hence, this particular value of “Area” can be

used as a cutoff in order to ascertain whether the upper layer has a higher stiffness than

the underlying layer(s).

The surface deﬂection at the center line of the circular plate may be calculated by

using Boussinesq’s solution for semi-inﬁnite half space by using equation (2-2) [17, 18].

2(1—v2)0'0 *a

 

0
Where;
E0 = The surface modulus (psi);
0}, = Peak pressure of FWD impact load under loading plate;
do =Peak surface deﬂection at center of load plate (inches);
a =radius of F WD loading plate;
v =Poisson '3 ratio of the semi-infinite half space

“2 " in above equation is a factor for uniform stress distribution [18]

The equation (2-2) can be simpliﬁed to equation (2-3) by assuming a Poisson’s

ratio of 0.5 for the half space.

* *
E0=(1.5 a 0'0) (23)
d0

 

The problem arising from an unknown stress distribution for equation (22) and
(2-3) may be solved by measuring the deﬂections at different distances from the center of
the load [17]. Measuring the deﬂections at several distances (as is the case with FWD)
also serves as a check on the assumption that the subgrade is a linear elastic semi-inﬁnite
Space. In that case the moduli calculated at different distances must be identical,

otherwise the subgrade is either non-linear or it consists of different layers.

19

 

E - = (2-4)
1 at:
’i’ di
Where;
E i = The surface modulus (psi);
00 == Peak pressure of F WD impact load under loading plate;
d,- =Suiface deﬂection at distance “i " from center of load plate (inches);
a =radius of F WD loading plate;
v =Poisson 's ration of the semi-inﬁnite halfspace
r,- =Distance from the center of the load where d, is measured.

Equation (2-4) can be simpliﬁed to Equation (2-5).

(0.84*00*a2)
Ei: (rt-Vi)

 

(2-5)

The E0 value calculated using Equation (2e3) represents a composite or effective
stiffness of g_ll the layers under the FWD loading plate. Based on Equations (2-1), (2-3)

and (2-5), the following equations were derived [15].

AF = (k2 ‘1) (2-6)

 

 

 

 

AF = Area Factor, i.e. the ”improvement" in AREA from 11.037 squared;
k1 = 11.037 (the Area when the stiffness of the upper layers is the same as that of the lower layers;
k2 = 3.262 (Maximum possible improvement in Area = 36/1103 7).

Equation (7) can effectively be used to approximate the relative stiffness of the

Upper (bound) layers in the pavement cross section.

20

ES: EO*AF*(k3)ll"/'1175l’1} 1 (2-7)

Where;

ES = Effective Stiffness of upper (bound) layers;

Ea = as deﬁned by Equations (3);

AF = as deﬁned by Equation (6);

k3 = thickness of upper layer/load plate diameter =

 

* a
2.4 REVIEW OF STATISTICAL METHODS

Previous studies that dealt with LTPP data (GPS and SP3 experiments) quantiﬁed
performance based on engineering criteria and expert opinion. For example, the

engineering criteria may include the rate of growth, severity levels and impact of distress

on the functionality of the pavement.

' Several statistical methods can and have been employed to establish performance
- criteria, to study the effect of design and construction features on pavement response and
performance. The statistical methods range from trend plotting to complex multivariate
analysis. A detailed statistical methodology for LTPP data analysis for prediction

modeling (performance models) has been laid out in SHRP-P-393 report [19].

The simpler statistical methods include Univariate and Bivariate analysis of data.
These methods include the determination of data statistics such as mean, standard
deviation and data frequencies. Simple histograms and box plots can also be generated to
determine data distribution. These methods also allow for determining the degree of
dependence between variables. The results of such an analysis can be graphically
illustrated. Such an analysis can also provide summary statistics such as the coefﬁcient

of correlation. Bivariate analysis can also assist in identifying outliers.

21

Hypothesis testing is a tool that allows one to determine if a speciﬁc numerical value is
equal to, greater or less than a speciﬁed number, or if the means of two sets of data are
equal or signiﬁcantly different. The mean response values for each group can be
determined and then compared using hypothesis testing for a certain level of conﬁdence.
If this relationship is signiﬁcant then the impact of the given factor on the response

should be further investigated.

Some of the multivariate statistical methods include:

Analysis of Variance (ANOVA)
Regression Analysis

Principal Component Analysis
Discriminant Analysis

The ANOVA is a tool that allows for better understanding of how the independent
variables inﬂuence the dependent variable. For example, the effects of different base
types or asphalt thickness on the amount of cracking (dependent variable) are required

candidates for such an analysis.

Regression analyses attempt to explain some dependent variable, y, in terms of
many independent (explanatory) variables, x ’s. The model (equation) can be either linear
or non-linear, and with actual, transformed, or interaction clusters of variables. The

model coefﬁcients can be estimated using best (least squares) ﬁtting techniques.

Principal component analysis is a multivariate procedure which allows for
simultaneously examining the data for both inﬂuential observations and collinearity. In
essence, it is a diagnostic/explanatory data analysis tool used in modeling to identify the

relationships between variables, and can help in deciding the reduction of the number of

22

variables (independent) so that these may be lumped into a fewer number of

factors/variables which explain the greater part of the variance in the data.

Discriminate analysis allows one to distinguish between two or more groups of
data. This is done by identifying which variables are signiﬁcant in classifying the data
into various groups. The procedure for predicting membership is to initially analyze
pertinent variables where group membership is already known. For example, groups of
observations can include one group of pavements with poor performance and the second
group with good performance. The method allows for determining which variables

discriminate between poorly and good performing pavements.

Cluster analysis is in some way similar to factor analysis as both of these
statistical techniques take a larger number of cases or variables and reduce them to a
smaller number of factors (or clusters) based on some sort of similarity those members
1 within a group share. The factor analysis is used to reduce a larger number of variables
to a smaller number of factors that describe these variables. On the other hand cluster.
analysis is more typically used to combine cases into groups. However, the statistical

procedures underlying each type of analysis and interpretation of the output are different.

Logistic regression is used often in the case where the outcome variable is
discrete (dichotomous or multi-nominal). The difference between logistic and linear
regression is reﬂected both in the choice of a parametric model and in the assumptions.
This method is based on the maximum likelihood method for determining the parameters
of interest. The interpretation of effects for various levels of the categorical variables

(independent) is very convenient in terms of the odds ratio when this type of model is

23

used. Logistic regression models are also very useful for discrimination analysis (of

various groups) when categorical variables are used as independent variables.

Survival Analysis can be used to study longitudinal data when survival time data
that measures the time to a certain event (such as failure, death or some threshold value
of distress) is the focus of interest. These times are subjected to random variations and
like any random variable, can be represented by a distribution. The distribution of
survival times is usually characterized by three functions: the survivorship function,
probability density function and hazard function. These three functions are
interdependent on each other, i.e., if one of them is known the other two can be derived.
The non-parametric methods deveIOped for survival analysis (Cox’s regression) are
capable of handling discrete and continuous independent variables effects in the
multivariate survival function. Thus this technique has the potential for analyzing

_ pavement time to failure given various design and construction factors.

The detailed discussion and description of the statistical methods used in this

research is presented in Chapter 5 under the research methodology.

24

CHAPTER 3 - DESCRIPTION OF SPS-l AND SPS-8
EXPERIMENTS

3.1 INTRODUCTION

This chapter describes Speciﬁc Pavement Studies (SPS) experiments 1 and 8, in
terms of their respective goals, experimental designs, and associated factors (design and
construction). A separate section is included for the Dynamic Load Response (DLR)

experiment which constitutes a subset of the SPS-l and -2 experiments.

3.2 STRATEGIC STUDY OF STRUCTURAL FACTORS FOR FLEXIBLE
PAVEMENTS—SPS-l

The effects of drainage, pavement structural parameters (thickness and stiffness of
asphalt concrete surface, and strength and thickness of base layers), and climatic factors
could be rationally accounted for in current design procedures for ﬂexible pavements.
However, some of these effects were developed through theoretical approaches because
of the limited available ﬁeld data. The in-service tests in this experiment will help
quantify the inﬂuence of these parameters on pavement performance and improve current
design procedures. Consequently, the experiment will help highway agencies in selecting
more economical design options for new and reconstructed ﬂexible pavements.

The SPS-l experiment is focused on the strategic study of structural factors for
ﬂexible pavements, and was intended to study the effect of speciﬁc design and
construction features on pavement response and performance. As the test sections in the
experiment are monitored since inception, the experiment provides an opportunity to
determine the relative inﬂuence of the key pavement design and construction elements

that affect pavement performance.

25

3.2.1 Experiment Design

Generally, a scientiﬁc study can be categorized as either an experimental study or
an observational study. This distinction is important because experimental studies
provide a much ﬁrmer basis for the establishment of cause-and-effect relationships one or
more explanatory factors and a response variable than do observational studies. With the
later one can establish only associations between the explanatory factors and response
variables, but not the causation. The experimental studies allow the researcher to control
effects of various treatments and difference in response between the treatments is only
attributed to the factors in the study.

Randomization of treatments to the experimental units is employed to neutralize
the effects of any unknown factors between the experimental units. However, if the
randomization is not possible then the difference in the experimental units can be handled
by using co-variates (variables effecting response other than experimental factors) in
order to detect the pure effects of factors under study.

The SPS-l trial is an experimental study, which aimed at ﬁnding the effects of
strategic structural and site factors on various pavement performances (cracking, rutting
and roughness) and response (deﬂections and strains). The site factors include: climatic
region (wet-freeze and wet-no-freeze, dry-freeze and dry-no-freeze), subgrade soil (ﬁne-
and coarse-grained), and trafﬁc rate (as a covariate) on pavement sections incorporating

different levels of structural factors. The structural factors include:

o drainage (presence or lack of it),

- asphalt concrete (AC) surface thickness — 4” (102 mm) and 7” (178 mm),

0 base type — dense-graded untreated aggregate base (DGAB), dense-graded
aSphalt-treated base (ATB) and a combination of both,

0 base thickness — 8” (203 mm) and 12” (305 mm) for un-drained sections; and 8”
(203 mm), 12” (305 mm) and 16” (406 mm) for drained sections.

26

The study design stipulates a trafﬁc load level in excess of 100,000 Equivalent Single
Axle Loads (ESALs) per year for the study lane [1].

The experiment includes 'four structural and two site factors with 2 (Drainage), 2
(AC surface thickness), 3 (Base type), 3 (Base thickness) and 4 (Climatic zones), 2
(Subgrade type) levels respectively. These levels for all six factors will render a total of
288 (2 x 2 x 3 x 3 x 2 x 4) runs. The number of runs indicates the possible treatment
combinations between the levels of all the factors considered in the experiment. This full
factorial design is shown in Table 3-1 for a single replication. However, due to economic
considerations, the total number of runs were reduced to 192 [288- (16+l6+16+48)].

This drop in the number of runs was accomplished by dropping some of the levels within
factors (16-inch base thickness in non-drained pavements and ATB/DGAB combination
for drained pavements).

Table 3-2 shows the modiﬁed experiment design. The reduction in the number of .
pavement sections seems to be the'only advantage by'drOpping some of the treatment
levels. However, this fractional factorial design will restrict the calculation of the higher
order interactions and certain effects will be confounded (can not be separated) within
each other. Table 3-2 also indicates that a total of 24 treatment combinations for
structural factors are required in each subgrade-climate combination. These 24
treatments represent different pavement design options. Furthermore, 12 designs were
assigned to two sites in the same subgrade-climate combination with an intention to
reduce the number of sections for each site and to maximize the participation of states in
the experiment. This partitioning of 24 designs (12 designs at two sites) will also cause
an increase in the variability in the observed performance of the pavement sections

because of the site related factors (materials, construction quality and measurement).

27

The fractional factorial design for the SPS-I experiment is shown in Table 3-3. The
overall experiment consists of 192 factor level combinations which consist of 8 site-
related (subgrade soil and climate) and 24 pavement structure combinations. The
experiment design requires that “48 test sections representing all structural factor and
subgrade type combinations in the experiment are to be constructed in each of the
climatic zones, with 24 test sections to be constructed on ﬁne-grained soil and 24 test
sections to be constructed on coarse-grained soil”[2].

According to the experiment design, twelve test sections were constructed at a
given project location (site). Each section is represented by either XX-OlOl to XX-0112
or XX-0113 to XX-0124, where XX denotes the state ID. Six sections have a target HMA
surface thickness of 4-inch (102 mm) and the remaining six have a target HMA surface
thickness of 7-inch (178 mm). Out of 12 sections, 5 have 8-inch (203 mm) base layer, 5
have a 12-inch (305 mm) base layer and the remaining 2 have a 16-inch (406 mm) base
layer. Also 2 test sections have dense-graded aggregate base (DGAB), 2 sections have
a‘sphalt treated base (ATB), 2 sections have a combination of ATB/DGAB, 3 sections
have permeable asphalt treated base (PATB) over DGAB, and 3 sections have ATB over

PATB. In-pavement drainage is provided only for sections with PATB as the base.

28

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

Table 3-1 Full Factorial Experiment Design Matrix

wﬂ 8
090.. 88888888888888888888888888888888888800
RT 2
N
C 11111111111111.1111lllllllllllllllllllllvw
F
F 111111111111111111111111111111.111111%
F
s
S
Am
MC .4../u4 m/ua._.lu4 MIN». m/ua.m/ ¢T¢T¢T¢T¢T¢Tu4mlm~mlmw Tm. m/ua. .:l.4 _./
H..m
T.
s
S
mm H n n n u .n n u n n n ..
u 2 6 n 2 6 u 2 6 u 2 6 n 2 6 n 2 no 1
BMW- 8 1 1 8 1 1|. 8 1 11 8 1 1 8 1 1 8 1 1 MOM
m T
m
1m
0
C

Base Type

DGAB

 

ATB

 

ATB/DGAB

DGAB

 

ATB

 

ATB/DGAB

 

 

Drainage

 

No

 

Yes

 

 

 

29

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

VN «N «N vN 4N VN «N «N
NS .25 5:28
N. N. N. N. N. N. N. N. N. N. N. N. N. N. N. N.
M. . . . . . . . . ..N
.2
M. . . . . . . . . a.
M. . _ . . . . . . ..N N. 55.5...
m . . . . . . . . a. ..
w . . . . . . . . ..N w
m . . . . . . . . ._.. ..
mo>
w . . . . . . . . ..N
.2
N . . . . . . . . a.
m . . . . . . . . ..N
..N. m<oo.m...<n.
w . . . . . . . . a.
w . . . . . . . . ..N w
w . . . . . . . . a. ..
w . . . . . . . . ..N
..N.
w . . _ . . . . . a.
@595...
w . . . . . . . . L w
w _ . . . . . . . a. ._
w . . . . . . . . ..N
..N.
me< 0z
w . . . . . . . . .... M.
m . . . . . . . . e. ._
w . . . . . . . . L
w . . . . . . . . v ..N.
. m<oa
w . . . . . . . . .e w
w . . . . . . . . a. ..
.85 U n. U a. U n. U n. $05.25. <23 $65.25. 3am on»... gum ummEEQ
as. ...zo .E “.23 ".3 .

 

 

 

 

 

 

 

 

 

meoumozﬁwom can x282 ewmmom EoEtomxm Etouomm .meotomi N-m DEE.

 

3O

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

31

 

 

 

 

ON. ON. ON. ~NN. ON. «N. .N. N. ..N
N: N: N: N: N: N: N: N: _... ._N.
NN. NN. NN. NN. NN. NN. NN. NN. ._N
... :. E :. .: :. .: .: ...~ ._N. 553:
O: O: O: O: O: O: O: O: :N
NN. NN. NN. NN. NN. NN. NN. NN. a. ._N 8>
NO. NO. OO. NO. NO. NO. NO. NO. _.N __O.
.N. .N. .N. .N. .N. .N. .N. .N. 2
NO. NO. NO. NO. NO. NO. NO. NO. __N
ON. ON. ON. ON. ON. ON. ON. ON. _... ._N. N20955:.
O: O: O: O: O: N: O: N: ..N
NO. NO. NO. NO. NO. NO. NO. NO. 2 ..N
NO. NO. NO. NO. NO. NO. NO. NO. ._N N.
N: N: O: O: N: N: O: N: .a __ NEON.
N: N: N: N: N: N: N: N: ._N O Em:
NO. NO. NO. NO. NO. NO. NO. NO. _... __
3. NO. NO. NO. NO. NO. NO. NO. _.N N.
O: O: O: O: O: O: O: O: a. ..
N: N: N: N: N: N: N: N: _.N N 9? 02
NO. NO. NO. NO. NO. NO. NO. NO. .a __
E. E. E. E. E. E. 3. E. __N N.
NO. NO. NO. NO. NO. NO. NO. NO. 3. _. NEON.
.O. .O. .O. .O. .O. .O. .O. .O. =N N
N: N: N: N: N: N: N: N: z ._
> x 3 > N. N N N. o n. o z s. . v. N
8.80 NEE 3.80 2.5 3.80 NEE oEmoU NEE $238.? $05.25. 09¢. ummm 092.35
Emmy: oz mNmmE mNmmE oz mNmNau. <2: 2mm .
55 NN.?
on... asm .NocoN oszzu 232:5 20:52...

 

 

 

ONEOE .MENQQ .5520me ~ .mmm. WM mEE

 

3.2.2 Discussion on the SPS-l Experiment Design

Fractional designs are effective for characterizing the joint effects of multiple
factors. However, the number of treatments, which is a product of the number of levels
for factor, grows rapidly with increasing factors. For example, an experiment involving
three factors each with three. levels, will involve 33 = 27 treatment combinations. One
way of economizing will be to limit each factor to two levels, which will reduce the
treatment combinations to 23 = 8. These two—level designs are extremely useful in
exploratory or screening studies where the objective is to identify the most important
factors from a larger set of potential factors. However, if many factors are to be studied in
a single experiment, even two-level designs with single replicate may become
overwhelming. For example, an experiment with 10 factors will involve 21°=1024
treatment combinations. Therefore, in such cases, a subset of the treatment combinations
can be chosen so that little or no information'is lost concerning crucial main effects and
low-order interactions. This chosen subset of the treatment combination is termed as
“fractional factorial design”.

In light of above discussion on the aspect of “how to reduce the number of
treatment combinations”, the SPS-l experiment design could have been designed as a
half (144 treatment combinations required) or quarter (72 treatment combinations
required) fractional factorial design. The detailed methodology for designing these 2k-f
CXperiments can be found in references [3, 4], where k denotes the number of factors and

f the fraction.

The SPS-l experiment has some inherent concems, because of the reduction of

levels for some factors. For example, the accurate effect of l6-inch base thickness can

32

not be studied as this level was dropped for un—drained pavements. Also, the pure effect
of ATB/DGAB base type can not be evaluated because there are no combinations of this
base type for drained pavements. It is important to note that the effect of drainage is also

confounded with PATB base type as these two effects are not separable within the SPS-l

experiment design.

It would have been more logical and useful to keep all the levels of base thickness
and base type in the experiment design and a systematical approach would have been
used for designing a fractional factorial designs. This approach would have rendered a
more economical and simple design with all main and two-way interaction effects. In
order to establish a clear cause-and-effect, the inherent variability between sites is
another concern mainly because of difference in trafﬁc, materials and construction
quality. To resolve this issue, it would-have been, more useful to repeat 24 design

combinations in representative sites with four (4) climate or eight (8) subgrade-climate

combinations.
3.2.3 Current Status of the Experiment (Release 17 of DataPave)

The SPS-l experiment includes eighteen sites with twelve sections each, for a
total of 216 sections located in all four LTPP climatic regions. The Wet-Freeze (WP) and
the Wet-No-Freeze (WNF) zones contain the majority of the sections. This is in line with

the Common wisdom that WP and WNF conditions critically affect ﬂexible pavement

Performance.

The geographical distribution of sites within the SPS-l experiment is presented in
Figure 3-1 . The full factorial design for the SPS-l experiment design requires that a

total of thirty six (36) similar designs be replicated across eight (8) soil-climate

33

combinations. The 36 designs were reduced to 24 designs in each soil-climate
combination making the experiment design a fractional factorial. However, it was
considered that the construction of 24 test sections at each site would require a greater
effort on the part of the participating agencies [2]. Therefore, to reduce the cost of
construction the experiment was developed so that only 50% of the possible
combinations of factors (i.e. 12 test sections) will be built at each site. The experiment,
designed in a factorial manner to enhance implementation practicality, permits the
construction of 12 test sections (0101 through 0112 or 0113 through 0124) at one site
with the complementary 12 test sections to be constructed at another site within the same
climatic region on a similar subgrade type [1].

The LTPP IMS data (DataPave 3.0) shows that the site populations within the
SPS-l experiment design are not equally distributed. This deviation is partly because of -
the cutoff values of precipitation and freeze index used for categorizing the ‘Vvet/dry” and
“freeze/non-freeze” climates. The current status of the factorial design, along with the
current distribution of sites in each climatic zone, is shown in Table 3-4. As these
deviations are expected to seriously affect the results of the analysis (the experimental
design is unbalanced), this issue will be further discussed in Chapter 4 under design
Versus actual construction. It can also be seen from Table 3-4, that there is no replication
aVailable for sites in DF zone for different subgrade types. Therefore, the subgrade
effeCts in DF zone can not be estimated. Similarly, the results may be seriously
hanlpered due to the small number of sites in Dry zones. A discussion on the current
status of the experiment for each site can be found in Appendix A1 of reference [5] .

Uniform construction guidelines were established for test sections within SPS-l

expel‘irnent designs. These guidelines are presented in the next section.

34

 

 

 

Figure 3,-1 Geographical location of SPS-l sites

Table 3-4 SPS-l site factorial — From DataPave 3.0

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

8“,?ch Weta Dryb
ype Designs Non Total
Freezec Freeze Freeze Non-Freeze
IA (19)
0101-0112 OH (39) AL (1) - NM (35)
L Fine KS (20) . 9
_ MI (26) LA (22) _
0113 0124 NE (31) VA 69 -
0101-0112 AR (5) FL (12) NV (32) -
Coarse DE (10) 9
0113-0124 WI (55) (T); ((33: MT (30) AZ (4)
Total 8 6 2 2 18
Nete:

3. Wet Regions — Average Annual Rainfall > 20 inches (508 mm)
b. Dry Regions — Average Annual Rainfall < 20 inches (508 mm)
0. Freeze Regions —— Average Annual Freezing Index > 83.3 °C-day (150 °F-day)

d. Non-Freeze Regions -— Average Annual Freezing Index < 83.3 °C-day (150 °F-day)

35

 

3.2.4 Construction Guidelines for SPS—l Experiment

The study of the SPS-l experiment has speciﬁc objectives; mainly the experiment
was designed to study the inﬂuence of design and construction features on the response
and performance of new ﬂexible pavements. Therefore the focus of the experiment is on
the main factors (HMA and base thicknesses, base type and presence or absence of
drainage). The designs were repeated across 18 states in order to study the effect of
different climates and subgrade soils. To study the speciﬁc objectives of the SPS-l
experiment, it is essential to control for other sources of variability which can mask the
effects of the main factors. These factors may include differences in construction quality,
material properties and trafﬁc levels across sites. Therefore, each SPS-l project had to
meet certain construction criteria. To approach uniformity across projects, there were
limitations on the methods and materials used in construction, as well as requirements for

testing and continued monitoring. .These guidelines are outlined below.

Construction Requirements
1 Construction requirements were provided in the “Construction Guidelines”
section of the SHRP-LT PP Specific Pavement Studies: F ive- Year Report[1, 6] . The
overall length of each section was required to be 183 m (600 ft) with 152.4 m (500 ft) for
monitoring and 15.25 m (50 ft) on each side for material sampling. The distance between
each of these sections had to be long enough to allow sufﬁcient space (transition) for
changes in materials and thicknesses during construction. The suggested length for these

transitions was 30.5 m (100 ft).

36

Subgrade Requirements

The ﬁnished subgrade elevations could not vary from the design by more than 12
mm (0.5 inches). This could be determined using rod and level readings taken on the
lane edge, outer wheel path, mid lane, inner wheel path and inside lane edge at 15 m (50
ft) intervals throughout the length of the project. Surface irregularities could not exceed
6 mm (0.5 inches) between two points in any direction in a 3.05 m (15 ft) interval.

Modiﬁers may be used to provide a stable working platform for construction but not to

increase subgrade strength.

Base Layers

Two types of bases are included in each SPS-l project —— drained and un-drained.
The drained bases include a permeable asphalt treated base with edge drains. The un-
drained bases consist of dense graded materials. Two types of dense graded bases were .
Speciﬁed for the sections without drainage. The un-drained bases were used in sections
101—106 and 113-118 and were deﬁned as dense graded aggregate base (DGAB), asphalt
treated base (ATB), or a combination of these two materials. The drained base was used
in sections 107-112 and 119-124 with a combination with DGAB and ATB base types.

The requirements for each base type are as follows:

Dense graded aggregate base (DGAB)

Minimum 50% retained on the No. 4 sieve.
Top-size aggregate was speciﬁed as 1.5 inches (38 mm).

' Less than 60% passing the No. 30 sieve and less than 10% passing the No. 200 sieve.
Liquid limit less than 25 and plasticity index less than 4 for fraction passing No. 40
S leve.

In L. A. Abrasion Test, the loss must not exceed 50% at 500 revolutions.
The compacted lift thickness must not exceed 200 mm (8 inches).
' The DGAB must be compacted to at least 95% of maximum density.

37

In-place density of DGAB should be determined prior to the application of an asphalt
prime coat.

The base surface must be primed with low-viscosity asphalt and allowed to cure prior
to placement of the asphalt concrete surface.

The ﬁnished DGAB elevations should not vary from design by more than 12 mm (0.5
inches).

Asphalt treated base (A TB)

The aggregate used in the ATE layer must meet the same requirements as the
aggregate for DGAB layer.

Asphalt emulsions should not be used in ATB.

Experimental modiﬁers were allowed only in the supplemental sections.

No recycled HMA was allowed in ATB.

Forthe Hveem mix design procedure, the following requirements were required for
the ATB:

Swell 0.7 mm
Stabilometer Value 35 min
Moisture Vapor Susceptibility 25

Design Air Voids 3 to 5 percent

For the Marshal mix design procedure, the following requirements were required for
the ATB:

Compaction blows ' ‘ . 50

Flow 3 to 5 mm
Stability ‘ “ 4.4 KN
Design Air Voids 3 to 5 percent

The maximum compacted lift thickness for the ATB layer should be limited to a
maximum of 8 and 4 inches (200 and 100 mm) for the ﬁrst and subsequent lifts,
respectively.

The minimum compaction requirement was 90% of the maximum theoretical speciﬁc
gravity for the ﬁrst lift and 92% for subsequent lifts.

The ﬁnished surface of the ATB base should not vary from the design more than 12
mm, as measured using rod and levels.

The base layer thickness should not vary from design by more than 6 mm (0.25
inches).

Perm eable asphalt treated base (PA TB)

The drained base was used in sections 107-112 and 119-124 with a combination

of D GAB and ATB base types. Each of these sections included a PATB layer with edge

drain 5 to permit water to drain out of the pavement structure. The requirements for the

PATB layer were as follows:

An asphalt emulsion was not allowed as binder for PATB base layer.
The gradation for the PATB layer should be within the following ranges:

% Passing

Sieve No.
38 mm(1.5 inch) 100%
25 mm (1 inch) 95 — 100 %
13 mm (0.5 inch) 25 — 60 %
No. 4 0 - 10 %
No. 8 0 - 5 %
No. 200 0 — 2 %
o More than 90% of the aggregate has at least one crushed face.
0 No recycled HMAshould be used in PATB.
0 Compaction should be performed by using static wheel roller applying 0.5 to 1.0 ton
of force per foot of roller width.
0 No portion of the PATB should be day-lighted.
HMA Layer
The HMA surface layers were required to meet the following minimum
requirements.

For the Hveen mix design procedure, the following requirements were required for

the HMA mix:

Swell (maximum) 0.7 mm

Stabilometer Value (minimum) 37 min

Air Voids ~ 3- 5%
For the Marshal mix design procedure, the following requirements were required for
the HMA mix:

Compaction blows 75

Flow 2 to 4 mm

Stability 8 KN

No recycled materials were permitted in HMA mixtures.

The aggregates should have a minimum of 60% retained on the No. 4 sieve with at
least two fractured faces, and a minimum sand equivalent of 45.

The asphalt grade and characteristics should be selected based on normal agency

practice.
The use of modiﬁers should be discouraged in the main sections.

Liﬁ thickness could not exceed 102 mm and compacted thickness of any single layer

had to be at least 51 mm.
Longitudinal joints should be staggered between successive lifts to avoid vertical

j oints.
All transverse joints should be placed outside the main sections.
The thickness of the HMA layer (surface and binder) should be within 6 mm of the

thickness speciﬁed by the experiment design.

39

o The as-constructed ﬁnished surface should have a proﬁle index of less than 10 inches
per mile (158 mm per km) as measured by the Califomia-type proﬁle-graph.

Shoulders

o The shoulders placed on these projects should have a minimum width of 1.2-m (4 ft)
and have the full pavement structure across their width.

0 If possible, the shoulders should be paved full-width with the surface course to
eliminate longitudinal joints. If not, then the shoulders should be paved such that the
longitudinal joint is at least 205 mm (12 inches) outside the travel lane.

Drainage Materials

Filter fabric (or geo-textile) was required on sections that included a PATB layer.

This was speciﬁed to prevent the clogging of the PATB layer due to migration of ﬁne

material from the subgrade. The ﬁlter fabrics used should meet the American

Association of State Highway and Transportation Ofﬁcials-American Building

Contractors-American Road and Transportation Builders Association (AASHTO-ABC—

ARTBA) Task Force 25 recommendations, which include the following requirements for

the geo-textiles:

In order to separate the base layer from the subgrade non-woven and woven geo-
textile materials that conform to Class A requirements should be used.

The geo-textile material conforming to Class B requirements could be used in the
edge drains.

Geo-textiles should be overlapped a minimum of 2 ﬁ (610 mm) at all longitudinal
and transverse geo-textile joints.

Filter fabrics should be installed in accordance with the manufacturer’s
speciﬁcations.

For the sections where the PATB layer was placed on DGAB, the ﬁlter fabrics
should extend around each edge drain and wrap around the outer edge of the
PATB layer.

Exposure time of the geo-textile to elements between lay down and cover should
be limited to a maximum of 3 days.

Edge drains were to be installed on sections containing a PATB layer to collect

Water draining from the permeable base. The requirements on these drains were as

follows:

40

0 Inside and outside edge drains should be constructed for crowned pavements.

0 Edge drains should be at least 3 ft (914 mm) away from the edge of the travel
lane.

0 The PATB was recommended for backﬁll around the edge drains; however, other
open graded materials could be used as backﬁll materials if approved.
Collector pipes (slotted) should be at least 3-in (76 mm) diameter.
Outlet pipes (un-slotted) should be rigid plastic pipes with a minimum diameter of

3-in (76 mm).

- Drainage pipes should be sized for the expected discharge determined as part of
design. '

0 Discharge outlet pipe should be placed at a maximum interval spacing of 76.2 m
(250-ft).

Material Sampling and Testing

Sampling and testing were required for each of the material used for the
construction of sections. The material characterization is necessary to evaluate the
differences between the as-constructed sections within a site and between different sites
within the experiment. These measured parameters are used mostly in the design
procedures as well as to assess important performance characteristics of the materials.

A general sampling and testing plan was created for use as a guideline[6]. These
guidelines were then used to develop the sampling and testing plan speciﬁc to each site.
These plans were created prior to the construction of each project and the location of each
sample was predetermined. The following types of samples should be taken from each
project:

Bulk samples from the upper 305 mm (12 inches) of the subgrade.
Thin-walled tube samples of the subgrade to a depth of 1.2 m (4 ﬂ).

Jar samples for subgrade.

Bulk samples for the DGAB.

Jar samples for the DGAB.

Bulk samples for PATB.

Bulk samples for ATB.

Bulk samples for the asphalt mixes used in the surface and binder course.
Bulk sample of asphalt mixes used in all mixes.

Cored samples for bound bases and surface asphalt layers.

41

In addition to these samples, bulk samples were to be taken for the asphalt
cement, aggregates and un-compacted HMA mixes. These samples were to be stored for
long term. A series of auger probes should be performed in the shoulder of each test
section up to a depth of 6 m. This allows for the determination of the stiff layer depth.
Finally, as part of the construction activity, nuclear density and moisture testing should
be conducted at the location of the bulk sampling areas for the subgrade and on the top of
each layer in every test section. The type and number of tests per layer are given

elsewhere [6].

Monitoring Requirements

The monitoring of the sections at each site includes several types of data. These
include distress surveys, deﬂection measurements, transverse proﬁles and longitudinal
measurements. Each of these measurements has different frequency requirements which

can be revised over time.

Distress Surveys

A distress survey was to be performed on each section within 6 months of
construction. A manual distress survey should be performed on the sections biennially,
With the exception of “weak” sections (2, 5, 7 and 13 in SPS-l projects where distress

surveys should be more frequent). The survey could be postponed by a year if necessary.

Deﬂection Surveys
Deﬂection measurements should be collected using a falling weight deﬂectometer
(FWD) from 1 to 3 months after the construction of the project. The deﬂection survey of

these projects is to be completed biennially except of the “weak” sections (2, 5, 7 and 13

42

.4-

in SPS-l projects where distress surveys should be more frequent). This testing also

could be postponed up to 1 year if necessary.

Transverse Profi l e
Transverse proﬁle measurements should be taken at the same frequency and at the

same time, as the distress surveys.

Longitudinal Proﬁle

Longitudinal proﬁles should be taken on the sections within 3 months after
construction. These measurements can be postponed up to 3 additional months. The
“weak” sections (2, 5, 7 and 13 in SPS-l projects) should be monitored every 6 months
but monitoring can be postponed up to 6 additional months. The other sections should be.

monitored biennially and can be postponed by 1 year if necessary.

Traffic Data
Trafﬁc data should be collected on each site. The current requirement states that
weigh-in—motion (WIM) data should be continuously collected on all SPS-l sections.
Continuous data collection has been deﬁned as the “use of a device that is intended to
operate throughout the year and to which the SHA commits the resources necessary to
both monitor the quality of the data being collected and to ﬁx problems quickly upon
determination of any fault” [l]. WIM devices are to be calibrated biannually. This level
of data collection is necessary to assess accurate trafﬁc loading measurements.
Clinatic Data
Each SPS-l site was required to install an automatic weather station (AWS). The

AWS should be located close enough to each of the sites to provide weather data that is

43

representative of the climate on each site. The equipment installed at these locations
should measure the following weather components:

Rain
Humidity
Wind speed
Temperature

All the data collected should be stored by a data-logger. The data should be
downloaded from the data-logger at least every 6 months.

In addition to AWS used to collect weather data, weather data should also be
obtained from the four or ﬁve closest National Oceanic and Atmospheric Association
(N OAA) weather stations. The data should be averaged using the weighting procedure,
with the weights based on the distance of the weather station from the particular site. The
data collected from NOAA stations should include information about the temperatures,

rainfall, wind and solar radiation levels.

44

 

3.3 STRATEGIC STUDY OF ENVIRONMENTAL FACTORS FOR
FLEXIBLE PAVEMENTS— SPS-8 EXPERIMENT

The SPS-8 experiment evaluates environmental effects in the absence of heavy
trafﬁc loads. The study examines the effect of climatic factors and subgrade type (frost-
susceptible, expansive, ﬁne, and coarse) on pavement sections incorporating different
designs of ﬂexible and rigid pavements, which are subjected to very limited trafﬁc as
measured by ESAL accumulation. Pavement structure includes two levels of structural
design for each class of pavements. Flexible pavement sections consist of 4” (102 mm)
and 7” (178 mm) HMA surfaces on 8” (203 mm) and 12” (305 mm) layers of DGAB,
respectively. Rigid pavement test sections consist of 8” (203 mm) and 11” (279 mm)
d‘oweled JPCP slabs on 6” (152 mm) DGAB. The study design stipulates the trafﬁc
volume in the study lane be at least 100 vehicles per day but not more than 10,000
ESALs in a year. The combination of study factors results in four possible section
combinations, two ﬂexible and two rigid. The ﬂexible and rigid sections may be
constructed at the same or at different sites. Table 3-5 shows the experiment design
matrix for SPS-8.

For ﬂexible pavements in SPS-8 experiment, the sections are identiﬁed as XX-
0801 to XX-0806 while rigid pavements are identiﬁed as XX-0807 to XX-0812, where
‘XX’ is the state code and ‘08’ stands for SPS-8 experiment. The sections with SHRP ID
that ends with an odd number have target HMA thickness of 102 mm or'PCC slab
thickness of 203 mm, while the others have HMA thickness of 178 mm or PCC slab
thickness of 279 mm thickness. In the section ID, an alphabet is introduced before the

SHRP ID in case a second site is constructed in the same state.

45

 

1|

3.3.1 Discussion on the SPS-8 Experiment Design

The SPS-8 experiment design for ﬂexible pavements is not a full factorial design.
Table 3-5 shows that there are 2 design factors (AC surface thickness and base thickness)
and 2 site factors (Subgrade type and climatic zone). There are 2, 2, 3 and 4 levels for
each of these 4 factors, respectively. Accordingly, a full factorial design involves 48 (2 x
2 x 3 x 4) treatment combinations (runs) for a single replicate. The experiment was
reduced to only 24 runs for economic and practical reasons. Consequently, the pure
effects of AC surface thickness and base thickness can not be studied from this study.
However, this reduction in the runs in SPS-8 experiment design seems logical in light of

the speciﬁc objectives of this study.

3.3.2 Current Status of the SPS-8 Flexible Pavements

Table 3-6 shows the presence of ﬂexible pavement sites in the SPS-8 experiment.
There are ﬁfteen (15) sites available for SPS-8 ﬂexible pavement sections with the largest
number of sites (7) located in the WP climatic zone. The distribution of test sections for
ﬂexible pavements according to SPS-8 experiment design is given in Table 3-7. In total,
32 ﬂexible pavement sections have been constructed in the 15 sites. There are a limited
number of test sections in the Dry zones with no pavements on ﬁne subgrade in DF or

active subgrade in DNF zones.

3.3.3 Construction Guidelines for SPS-8 Flexible Pavements

Construction guidelines were provided to ensure uniformity and consistency
among the test sites. The requirements for preparation and compaction of the subgrade

for ﬂexible sections are the same as for the SPS-l experiment.

46

 

 

 

 

 

 

 

 

gnome cm; .9360 03 v zoo:— wENooi _N===< owEo>< l Neewox 33....-.52 .o
93-..... on C QRUG m.mw A .89: 3.38... _m==:< omEo>< l 229M 032". .o
NEE wow. 85:. on v .5223. 72.5.... omEo>< 1| Ncoﬁom be .n
NEE wen. 3:2: on A :ﬂ: 3. _m:::< owﬁo>< 11 2.2»; .03 .m .202
m. m m m N .38.
.NN. .3
N G. <0 mm... .NNNNN. AN. 02 .NN. <3 3:8
SN. >2
.ON. :0
a... OS .ON. .2
2:
N .NN. .22 6... am NON. ms. .ON. 03. ....
AN. m<
QNUOHH—IEOZ DNOULn— ONOOLN—uccz DONOOLn—
.soN . O 25 82.3%
to «.03

 

 

 

 

 

N.
NEON 0.88:0 one 090 25.ng .3 Noam 35:622. o

3.3 3%.... 8.3.5.5 ON 22¢

:8 2.83:... 35.01.98.600 .0
:8 02:33 352905; . 1,.
3:3 2.350 9 3:22 on». $5.25 .0 03.3083 60.. .228 :8 383.5 92.3.. .<

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

x x x x x x x x x x x x o _ _
23.

x x x x x x x x x x x x o w

x x x x x x x x x x x x N. N.
2.:on

x x x x x x x x x x x x w v

D m < D n. < O m <. D m < $05.35. mas—2.3. on»...

mZQ NE ”.23 E5 omwm UUQU< EoEoZE

 

 

 

v..O_....>. ONE: Evacuaxm NN.: N-N sea

47

:8 o... .0 38:8 2.. chEE 22.08 .0 8888 8.86:. .o: comm .090 3328.8. w::_o3m ..o 035838-80... 856 on :8 :8 02.3...
. .88 88.8%.. outﬁwomcoo um<OD 8.92

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

. _ o _ o N . N _ . v o m N. N. 03.on
. _ o N o _ N _ _ m N N w v .
4N NN NN .N ON w N. N. O. N. z ... N. .. O. O O N O N v N N . €2.25 $2.25
8.80 2.. a. 8:2 8.80 8.... 2.8.... 8.89 02E o>_.o< 8.80 8.... .os..o< $05.2”... $05.02.... on»...
. 0821.102 082... 082 “To Z 082... m<O.O Suﬁsm .8258:
be 33
on»... unﬁwosm one 2882.th 8.862 85.02% .8808:

 

NEON ozmﬁzobmb ooﬁwnzm .:wmmoo .3 9.2.08 NEoEoZNNN 2&ch Wmmm mo 83.5.85 Wm 28...

48

The construction guidelines stipulated the use of DGAB; the requirements for the
materials and construction of the base layers are the same as those of the SPS-l
experiment. Similarly, the guidelines for the materials and construction of asphalt layers
are similar to the un-drained sections included in the SPS-l experiment, which require

DGAB.

3.4 INSTRUMENT ED SPS TEST SECTIONS

The Strategic Highway Research Program (SHRP) has included two projects with
instrumented test sections as part of the SPS-l and SPS-2 experiments. This subset of
sections is referred to as the Dynamic Load Response (DLR) experiment. The sections
are located in Ohio and North Carolina. The Ohio test sections include both ﬂexible and
rigid pavements while the North Carolina test sections include only rigid pavements. The
objective of this experiment is to support the development of mechanistic-empirical
design procedures for ﬂexible and rigid pavement systems. More speciﬁcally, it can be.
used to investigate the relationship between pavement response and performance, and to
validate pavement response and performance prediction models.

In addition to standard FWD, proﬁle and distress measurements, pavement
dynamic response parameters are being measured in these test sections including:

Vertical deﬂections in the surface layer, base and subgrade;
Horizontal strains in the pavement;

Vertical pressure at layer interfaces; and

Joint opening in PCC pavements.

The seasonal parameters being measured include:

Temperature within the pavement layers including base and subgrade;
Frost depth in base and subgrade;

Soil suction in the subgrade;

Water table elevation; and

' Moisture in the subgrade.

49

In addition to measurements within the pavement system, the loads being applied on the
pavement are measured using weigh-in-motion (WIM) scales. Several series of controlled
vehicle tests at different speeds and non-destructive load tests have been conducted to

measure the pavement response.

3.4.1 SPS-l Experiment Sections

The instrumented ﬂexible pavement sections are located in the SPS-l site Ohio
(3 9). These sections were instrumented with strain gauges, pressure cells and multi—depth
deﬂectometer (MDD) to conduct the controlled loading experiments.

The experiment targeted four core sections for the installation of sensors to
monitor dynamic pavement response during controlled vehicle testing. These sections
include 39-0102, 39-0104, 39-0108 and 39-0110. Tests were to be performed with a
single axle and tandem axle dump truck. Theqrear axle on the single—axle t1uck was
loaded to approximately 18 kips (40 kN) and 22 kips (49 kN) while the total load on the
rear axles of the tandem-axle dump truck were 32 kips (142 kN) and 42 kips (187 kN),
respectively. Both trucks ran over the instrumented sections at 50(30), 65(40) and 80(50)

km/hr (mph) in the morning and in the aﬁemoon.

Experiment Setup

The details of the instrumented ﬂexible pavement sections are given in Table 3-6.
Tests were conducted in the morning and in the afternoon to gather information on how
temperature differences in the pavement layers affect response. Table 3-9 shows the
instrumentation details of all the strain gauges and Linear Variable Differential
Transformers (LVDT) for each instrumented section. This information is taken from
Report No. FHWA/OH-94/019 by Ohio University, as this data was not available in the

DataPave Release 17.0.

50

Table 38 Details of instrumented sections for ﬂexible pavements

 

 

 

 

 

 

 

 

 

 

Section ID HMA Base Thickness Drainage Comments
Thickness (inches) / Base
(inches) Type
39-102 4 12 DGAB N0 Strain gauges at 4”
39-104 7 12 ATB N0 Strain gauges at 7" and
19”
39—108 7 4 PATB Yes Strain au 7”
8 D GAB g ges at
39-1 10 7 4 ATB Yes Strain gauges at 7” and
4 PATB 11”

 

 

Note: DGAB — Dense graded aggregate base, ATB - Asphalt treated base, PATB - Permeable asphalt treated base

Table 3-9 Instrumentation details for all the SPS-l sections in Ohio

 

 

 

 

 

 

 

 

 

 

 

86:3“ Strain Gauge Designation Location LVDT Designation
DYN7 — Transverse
DYN8 — Longitudinal A“ .Stram gauges LVDTl — Deepl
DYN9 _ T are installed at the 2
39-102 ransverse bottom of AC, 4" LVDTZ - Shallow
DYNlO — Longitudlnal deep from the LVDT3 — Shallow
DYNl l - Transverse surface LVDT 4 — Deep
DYN12 — Longitudinal
DYN 10 - Transverse 2:111:21ng 1 5
gm}; _ 'Ill‘ransvegse 1 deep from the
— ongltu lna surface _
DYN14 — Transverse thT; _ 1831363
39104 DYNlS — Longitudinal LVDT3 _ $112113:
DYN16 — Longitudinal These three, snail? d LVDT“ — Deep
_ ~ . 1 gauges are lnsra e
a
g ATB, at 19” deep
from the surface
DYN 10 — Transverse
DY-Nl l — Longimdinal A“ Strain gauges LVDTI _ Deep
DYNIZ — Transverse are installed at the LVDTZ _ Shallow
39-108 DYN13 — Longitudinal bottom of AC, 7” LVDT3 Sh 11
DYN14 - Transverse deep from the LVDT4 — a ow
DYN l 5 — Longitudinal surface ‘ Deep
DYNlO — Transverse DYNI 0 to DYN 15
DYNll —Longltudlnal are located at
DYN 12 - Transverse bottom of AC, 7..
DYN13 _ Longlmdlnal deep from the LVDTI _ Dee
DYN 14 — Transverse surface LVDTZ _ Sb 1‘:
39-110 DYNlS — Longitudinal a 0‘”
These three strain LVDT3 _ Shallow
DYN16 _ Longitudinal gauges are installed LVDT4 _ Deep
DYN17 — Longitudinal at the bonof,“ 0f
DYNlS— Longitudinal ATB: 3‘ 11 deep
from the surface

 

 

8“ ~ 4
he deep referenced LVDT lS anchored at 10 ft depth from the surface. ' The shallow referenced LVDT is
anchored at bottom of base layer for each section from the surface, Source: Report No. FHWA/OH~94/019

51

CHAPTER 4 - DATA AVAIABILITY AND EXTENT

4.1 INTRODUCTION

This chapter presents a summary on the available data and extent of various
performance measures for SPS-l and SPS-8 experiments. For this study DataPave
(Release 17, January 2004) was the primary data source. However, data from previous
releases and other sources were used only to supplement the level E data from Release
17. The construction information of each site was extracted from the construction
reports. The essential data required for this study can be broadly classiﬁed into following
categories:

0 Site Information: site information, construction issues, climatic and trafﬁc
data.

0 Material Data: Material type and properties for various bound and un-bound
pavement layers.

- Pavement Structure: Layer type and thickness information and other design
features such as lane width, shoulder type and dowel bar diameter etc.

0 Monitoring Data: Longitudinal and transverse proﬁles, distress and

deﬂections.
- Dynamic Load Response Data: Response data for instrumented sections.

4.2 IDENTIFICATION OF DATA ELEMENTS

The relevant variables contained in the SPS-land SPS-8 experiments can be
divided into: (1) dependent variables, and (2) independent variables.

The dependent variables are those used to describe pavement response and
performance. Measures of pavement response are those measures that do not cumulate
with time. The bulk of pavement responses in these experiments are surface deﬂections
from Falling Weight Deﬂectometer (FWD) testing. For ﬂexible pavements, FWD testing

is conducted in the wheel path and outside the wheel path. Other pavement responses

52

collected in the SPS-l experiment include strain data and vertical deﬂections at various
depths. These measurements are only. available from the instrumented sections in the
Dynamic Load Response (DLR) experiments.

Measures of pavement performance are those that cumulate with time (e.g.,
alligator cracking in ﬂexible pavements). These are collected using both manual and
automated surveys.

The independent variables are those that describe design and construction factors.
These can be divided into: (1) main variables, and (2) exogenous (or confounding)
variables. Main variables are those used to specify the design matrices of the respective
SPS experiments (e.g., base type). Whereas, the variables that have potential impacts on
pavement response and performance but are not controlled in the experiment design were
considered as exogenous variables. Exogenous variables that are independent of the main
experiment variables are the actual'cumulative trafﬁc (KESALs) and age. All other
exogenous variables are associated with the main design and construction variables.
These include: (1) material properties of the various pavement layers, Which constitute
the structural factors in the design matrix, and (2) climatic factors, which describe the
four climatic zones in the matrix. Figure 4-1 shows the summary of main variables
(independent) along with their levels considered for explaining the dependent variables
(performance measures) for the SPS-I experiment.

Table 4-1 lists the relevant independent and dependent variables identiﬁed for
ﬂexible (SPS-l and SPS-8) pavements. After the identiﬁcation of the relevant variables, a
database was developed for this research. The development of database is brieﬂy

discussed next.

53

moist? Eoﬁﬁﬁobﬁ can Engage mo bmﬁﬁzm TV 235

 

 

 

Yﬁ mo>o
OZ.

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

" n
. .
. .
u n
. .
m H umaEEQ W m
u. ..................... u m m
. . . .
. u u u
m N azo .J . " rﬁ 309m: . n
u .5 . . u m: . u
m “.23 . m " m<oo . u
. .
m m3 . m m 09C. 345 n
. ocoN 235:0 . _ ”Illll
" r r C all: _ .m A83a2a>
" a u u :8 . " Eat—536:6
. _ .
n 3800 o n u _ an. o " 9.30am
" 2E . n u . an o N 5293350 M5
“ 0%... unswnzm " " 3.05.35. 8am " Q :wmmon—
. .
u u - A u
. . .
u ..................... u u u
r i\ u H 2&- O u
< u an o u
m _ $2225. 3. m
282$ 8mm m n
_ .............. _. ............. n
+ + + + + 4
3:5 £20883 AEV AEV AEV A863
3.3. 8:09:06 $5.25 €55 mac—oenu m32-w=§oﬂo m3-w=_onU 32 mac—080
Ea 28:8on 3: .mmocnwsom 530 5m 8555;. 359538..— EEvBEcoA Lousy—2

 

923.8% Q mucus—Stem 2.25.3.6 833...; 2.2.5.5:

Table 4-1 Categorized list of variables for ﬂexible pavements (SPS-l and SPS-8)

 

Factor

Factors

 

Environmental Factors

No. of days with Freezing Temperature

No. of days with temperature>32°C

Annual No. of days with precipitation

Annual No. of days with high precipitation

Avg. Annual No. of FT cycles

F1, Degrees-Days

Avg. Annual Precipitation

Environmental Zone

Avg. Max Temperature, °C, Avg. Min Temperature, °C, Avg. Temperature
Range, °C

 

Asphalt Concrete
Material Properties

AC Grade

Target AC Thickness, mm

Thickness deviations, mm

AC Back calculated Resilient Modulus

AC Indirect Tensile Strength aﬁer MR Test, kpa

AC Indirect Tensile Strength Prior to MR Test, kpa

AC Instantaneous Resilient Modulus at 5, 25 and 40 °C, MPa
AC Total Resilient Modulus at 5, 25 and 40 °C, MPa

Bulk Speciﬁc Gravity of AC Mix

Water absorption for AC mix aggregate

AV%, AC%, AC mix gradation (all sieves) and AC viscosity at 60 °C

 

Aggregate Base
Material Properties

Target base thickness, mm

Thickness deviations, mm

Type of base (GB, TB, PATB)

Granular base Compaction (Max. density and OMC)
Base back calculated resilient modulus

Avg. Lab based granular base resilient modulus
Base gradation (all sieves)

Atterberg Limits_(LL, PL, PI)

 

Subgrade Material
Properties

Subgrade soil type
Subgrade Compaction (Max. density and OMC)
Subgrade back calculated resilient modulus

K1 ,K2 and K5 parameters from the resilient modulus testing for subgrade
Avg. Lab based granular base resilient modulus

Subgrade gradation (all sieves)

Atterberg Limits (LL, PL, PI)

Embankment heights (cut or ﬁll)

 

Trafﬁc/Age

Cumulative Annual Trafﬁc in KESALs
Average Annual Trafﬁc in KESALs
Age, Years

 

Performance

Alligator Cracking (fatigue)
Transverse Cracking

Longitudinal Cracking in WP and NWP
Bleeding

Raveling

Roughness (IRI)

Rutting

 

Response

 

 

Deﬂections
Various Deﬂection Basin Parameters
Strains (DLR)

 

 

Note: The variables in bold are the potential main factors (independent variables) and
performance/response (dependent variables). The variables in italics were considered as exogenous

factors.

55

 

-I-I

The data used in this study are “Level B” data from the IMS database (Release 17.0) for
SPS-land SPS-8 experiments. All data were extracted from the Release 17.0 CD. The
DLR data contained in the DataPave 3.0 database is insufﬁcient and/or inadequate for the
analysis.

For the purpose of this research, the data available along with its type is shown in
Figure 4-2. The ﬂowchart describing the process of data extraction from the DataPave
Release 17.0 is shown in Figure 4-3. The database has been set up such that the linkage
between different data elements is preserved. This was done using ACCESSTM,
EXCELTM and SPSSTM software. This relational database allows for describing the data
in different ways by combining various factors according to the speciﬁc objective of the
particular analysis at hand. Tables and ﬁguresproduced and presented in the data
availability section for all experiment designs are example outcomes of this data
structure.

For cases where multiple data values were available for a data element, the values
were averaged to obtain a unique value. For example, IRI values were averaged over ﬁve
runs for each section and for a particular date. Deﬂection measurements were averaged
over the length of each section for a particular date.

To complement/cross-check the inventory data available in Release 17.0, construction
reports (in the form of PDF ﬁles) for all sections within the SPS-l and -8 experiments

were obtained.

56

 

8% SEE
22.0.62 0
3% e832
2238805. o
9538
uszzu .
A5“: 3.53::
82. 8:8 25,—.

 

 

 

 

_
_

 

 

 

 

838% main 5 Saw Engage («o unﬁt miv Emmi

 

E028
2:562 o
8%
x_E=2%< a
Sun
38 ﬁwcozm o
m:E__w._o€u:< o
2"“:an
L8 coca—GEO o
35 2.39
32. .35...

 

\/

4.“:an cozosbmcou

 

05 68.: 22333
8:35.50

 

 

 

g
_

 

 

 

 

 

.3831
22323.50 o
BEEEm .

82.58%
934 0:2 o
mctozcoE o
$3.5 3:E—Se
:3» .:E 3<mm

 

 

 

 

 

 

_
_

 

 

.8M 0

2:... ucﬂmnzm .

owaEED -

on»... 3:832 o
«353:;

L995 .

3.:— 23m.

«use near.—

 

 

 

 

_
_

 

 

 

Sac wean—ox
Eon—cobém

8mm
noes—om £3532

«SD 058%

 

 

A5331
@525.

UP

352-“:
$5.04
9.22:0
SSE—2 o
Antwa— 35:55
5% 328 25...

 

 

 

 

 

_
_

 

 

 

 

%

a

A

 

8mm DEER:

 

 

oucmetoﬁom
“5:5me

 

 

 

omSSmQ
no Eng—2300

 

*

a

57

 

 

 

 

Select the states,
experiment type
(SPS-l or SPS-8)

 

 

   

  
       

Database

Data Type?

 

 

Performance Data
Monitoring Module
0 Cracking

- Rutting

- Roughness, IRI

- Faulting

o Pumping

- Spalling

Exploration and
Extraction module

 

 

 

  

 

Response Data
Monitoring Module
0 Deﬂections
0 Strains

 

 

 

 

 

 

Climatic Data

Climate Module

ITemperatures related

data

0 Precipitation and
moisture related data

 

 

Material

Properties/Inventory

Testing/SP5 Specific

Module

- Thicknesses

- As built type and quality
of materials in various
pavement layers

0 Laboratory testing

 

 

 

 

 

 

 

 

 

Select the
required Field
in each Table

to export the
necessary data

 

 

Traffic Data

T raﬂic Module

0 Historical trafﬁc

ESALs

o Monitored trafﬁc
ESALs and axle load

 

\ spectrum

 

 

 

Export the selected data to Excel or

Access

 

 

Figure 4-3 Data Extraction Process Flow Chart

58

 

 

 

 

The construction reports were reviewed for the purpose of obtaining additional
detailed information on construction and design features. They also include problems
encountered during construction of the SPS pavement sections. Some of these problems
have been highlighted in this Chapter. The reports were useful in conﬁrming/disproving
conﬂicting information as well as identifying and explaining some anomalies in the

database.

After extraction of all relevant data elements and building the analysis database, the

data were reviewed to determine:

0 the availability of main (design and construction) factors and exogenous
(confounding) factors in the identiﬁed data element tables,
the availability and extent of response and performance data,

0 the variability of response and performance measures, and
the variability of the main and exogenous factors.

The sections below present the details of the information on the availability and

extent of these variables for SPS-land SPS-8 experiments, respectively.

4.3 DATA AVAILABILITY IN SPS-l EXPERIMENT

Table 4-2 presents the overall availability of the relevant data elements within the
SPS-l experiment in Release 17.0 of DataPave. This includes both the main design and
construction factors as well as other exogenous factors such as trafﬁc, material, and
environmental data. The data is presented as a percentage of the total number of sections
for the main factors, while for the exogenous factors, data is expressed as a percentage of
the tOtal number of sites (states) included in the experimental design matrix. 5

The table shows that the data availability for the main factors is high, while that of

the eXogenous factors is somewhat lower. In particular, the availability of trafﬁc data

59

(61% for monitored data and 50% for estimated/historical data) is lower than expected
and should be improved. It should be noted that trafﬁc estimates from construction
reports are available for all but one site [MI (26)]. The availability of relevant data

elements is discussed in detail in the subsequent sections.

4.3.1 General Site Information
This section of the report presents the summary of the site identiﬁcation and

location, construction report availability and important dates associated with each of the
SPS-l projects. Also the details of other factors such as climate and trafﬁc, which pertain

to a particular site in the SPS-l experiment, will be discussed in this section.

Construction Reports
The construction reports have been prepared for each site by the supervisory

consultant on the project. These documents contain the details of the construction
process from conception to the completion. In addition, these reports presents
information on the geometric layout of various sections within a site, construction issues
(deviations from the guidelines, if any), trafﬁc, environmental conditions during the
construction and material quality control data. These reports are available for all the sites
in the SPS-l experiment except MI (26). A summary of the construction issues at each of

18 sites is given in Appendix A1 of reference [1]. These construction issues can be

helpful in explaining any poor performance at a particular site.

60

Table 4-2 Summary of SPS-l data elements availability

 

 

 

 

 

 

 

 

 

Data Category Data Type Data Atjﬂabrhty,
0
Construction Reports 94
Climatic data
Virtual Weather Station
Annual Temperature 100
Annual Precipitation 100
Automatic Weather Station
Site Information Monthly Temperature 83
Monthly Precipitation 83
Traﬂic data
Trafﬁc Open date 100
Estimated ESALs 50
Monitored ESALs 61
Axle Load Spectrum 72
Asphalt Layer
Core Examination 99
Bulk Speciﬁc Gravity 89
Max Speciﬁc Gravity 42
Asphalt Content 56
Asphalt Resilient Modulus 15
Penetration 49
. Viscosity 48
Matenal Data Asphalt Speciﬁc Gravity 47
Aggregate Gradation 56
Fine Aggregate Particle Shape 26
Layer Thickness 100
Unbound Base Gradation 20
Subgrade
Subgrade Gradation 44
Atterberg Limits ‘ 56
Subgrade Modulus 51
Layer details
Type . 100
Representative thicknesses 100
Pavement Structure Constructed thicknesses 94
Shoulder information
Type 86
Width 86
\_ Thickness 86
FWD data
Deﬂections 100
Temperature at Testing 99
Monitoring" Backc'alculated Moduli 6
Manual Distresses data 100
Longitudinal Proﬁle (IRI) 100
_ Transverse Proﬁle (Rut Deptli 100
Note; Data is said to be available for a section even if it is available for one survey.

61

 

Climate Data
The climate data were essentially used in deﬁning boundaries between various

climatic regions. The average annual rainfall for each site is considered as
discriminating variable between “wet” and “dry” regions, whereas, average annual
freezing index is used to locate each site in “freeze” and “no-freeze” regions. The climate
data is available form two sources in LTPP database— Automatic Weather Stations
(AWS) and Virtual Weather Stations (VWS). The AWS data are collected by a weather
station installed at each of SPS-l experiment site. AWS data for three sites; NV (32), OH
(39) and W1 (55) are not available in the Release 17.0 of the database. The VWS data are
collected from the existing weather stations in the vicinity of a speciﬁc site in the SPS-l
experiment. Climate data for all the sites are available from VWS. Therefore, in

DataPave, the climate data from VWS have been used to classify each site in a particular

Climate region/zone.
Trafﬁc Data
Heavy truck trafﬁc plays a vital role in determining the level of performance in

ﬂexible pavements. The trafﬁc data in terms of ESALs per year was obtained from
different sources. These sources can be summarized as;

0 IMS Database: The LTPP IMS database contains trafﬁc data in the following

forms;
0 Monitored Data—data obtained from weigh-in-motion equipment

installed at each SPS-l site.

0 Estimated Data—data obtained from the DOT’s based on their best

estimates from the previous history of the highway section.

62

lo Axle Load Spectrum—data obtained from the axle weight data; this is

essentially similar to monitored data.

The laCk of trafﬁc data for various states in the LTPP database has given rise to

the quest for reasonable trafﬁc estimates for the missing states. Therefore, other sources
were explored, including:

0 Construction Reports—the estimated design ESALs were taken from

construction reports for all the states in the SPS-I experiment.

- FHWA VTRIS database—this was used for estimating the average truck factors
for each site, once the ADTT is known from the construction reports,_the ESALs

per year were estimated for a particular site.

0 Previous Studies—the available studies on SPS-l [2, 3] were also used to extract

traffic information.

Finally, the ESALs per year were estimated by combining the information from
all the sources and conﬁdence levels were assigned to the quality of available trafﬁc data.
Table 4-3 summarizes the trafﬁc data availability for all the states within the SPS-l
experiment. The importance of trafﬁc data cannot be ignored in pavement design and

analysis; however; in this research only the trafﬁc estimate is required to neutralize its

eﬁ‘ects between different sites.

Trafﬁc opening date is the date on which a newly constructed project was opened
to tI‘afﬁc. This data is available in the database for all sites. The age of a section has

been calculated using trafﬁc opening date and corresponding last survey date.

63

85.5 as 33228 8.5 a ”202

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

2528 58 E... Lo 25; :82 .82 G. can a .2 .2 . - a: .i - - 2 :5 .5282;
3.? Based 38$ :9: RN :3 8a 5. an. o: - a: .to am : <> gee;
<35 33 8n - - 8o. 82 - 2 82 - - - e. V? .22.?
Eat a... 9% Sue 3.2: :09: :ozozbmcg wEmDn
e83. .250 3:0 33 SN - - am an - ex ea - - - 9. V5 .2225
$528 L=8 acid as? .5622 .3: can s: K can as. on - an ex - com a :c .220
:25. .acoo 283 an: a: some R. 9.. an or 8m 8m mm. ca 5.. mm 22 .8.on 3oz
2528 use an... Lo 2.? :32 :9: m: as a: an E ea one 8m 8m me. mm mm >z 5352
2228 as E: do as? as: :9: n: 9e: a 2. m: 2: a: a: E a: _: 2 m2 .Saaaoz
2528 son a: do as? as: .82 5 s2 8 E E - - a: 5 - - 2 :2 .2552
52.8%. 23 o_a< Ede 3:0 .82 a: son an m: m: E - - a: - R an .2 gaze:
tonne ago :5 33 3m - - an an . am an - - - 2 <4 .2253
2228 as E: do as? :82 E: as s: 2 NNN ”mm on - Mom oom :.~ SN om mo. .aaax
8.2.8:. assay a 3:; EB: .82 a saw 3 N: o: o: o: c: m: E .2 2 <_ 432
2528 L3,. an: do as? 55o: :2: a... so a. as. 3:. one .83 can we. - a: 2 d deco:
Bases a; canoe .aaooao :32 33 an an. a: an 2: 2a. am am - a; - 2 m5 .eaasoo
Ea eoaoz .aaoo 9589 a Base 3 oaamowaaﬁwm on: man to? 3 am at. as. o: a: we. .30 N2 m e< .3232
2528 58 E: .3 25; so: an: E as: a :N as on n: n: of E as a N< .2532
:23 582.28 so: :33 33 SN - - SN 5 - a 2m - - - _ a< .33.“?
S a w
3.38mi m m... .m m u ..o .l w w r m a W m. . meow 25m
P m W 32.3 ago an:
Seesaw meE=m Lama emu S<mm¥

 

 

 

 

 

 

SEE 2% egg 3mg 3 LE

 

64

4.3.2 Material Data

The data pertaining to various material related properties of various pavement
layers in the construction of each pavement section have been categorized in material
data. The data used in this research mainly include the material properties of the
subgrade soil (passing #200 sieve and Atterberg Limits). These data were used to verify
the subgrade soil classiﬁcation (ﬁne or coarse). However, in Release 17.0 of the
DataPave this data is only available for 44% and 45% of the sections for soil gradation
and plasticity index respectively. Therefore, the materials code available in the materials

ﬁeld was used to get the soil type for each section within SPS-l experiment.

4.3.3 Design versus Actual Construction Review

The SPS-l experiment is based on the fractional factorial design i.e., all the
combinations between levels of various factors were not taken in the design factorial.
However, the design matrix was populated with equal number of sites within each
climatic zone. To ascertain the homogeneity of the planned experiment with actual
sections in the ﬁeld, in this section the site and structural factors will be compared
between as-designed versus as-constructed. First a brief discussion on the construction
issues will be presented, and then the deviations in the site and design features within the

SPS-l experiment will be presented.

Construction Issues

The construction guidelines as discussed in Chapter 2 were speciﬁed for each site
Within the experiment. However, there were some deviations and construction issues
related to the some of the sites. This information was obtained from the construction
YEports. A brief site wise discussion on construction issues can be found in Appendix A1
Of reference [1]. Some of major construction issues which may have adverse effects on

65

the pavement performance for some particular sites in SPS-l experiment are summarized

below.

For SPS-l site in Kansas [KS (20)], it was mentioned in the construction report

that:
o The contractor experienced several problems during construction, many of
which were caused by the weather. The area experienced much higher than

average precipitation during spring 1993, resulting in delays and a wet
subgrade. To dry out the subgrade, the contractor was allowed to incorporate

ﬂy ash.
- During the FWD testing, high deﬂections were measured in the base in some

areas.
0 There was also segregation in the mix; these problems were “corrected” with
adjustments in construction methods.
Similarly for Texas [TX (48)], it was found that most of the sections prematurely

failed in rutting [4]. This rutting was attributed mainly to asphalt layers because of

following reasons:

0 Excessive asphalt content in the t0p layer.
0 Change in the gradation of the aggregates without modifying the asphalt mix.

Site Factors
The SPS-l experiment design stipulates that a total of twenty four (24) similar

designs will be replicated across eighteen (18) sites in the US. The experiment, designed
in a factorial manner to enhance implementation practicality, permits the construction of

12 test sections (0161-0112 or 0113-0124) at one site with the complementary 12 test
sections to be constructed at another site within the same climatic region on a similar
subgrade type[5]. Table 4-4 lists the intended sites in each subgrade type-climate zone
Combination within the SPS-l experiment [2, 3].

However, the LTPP IMS data (DataPave 3.0) shows that the sites within the SPS-

1 experiment design are not balanced. This deviation was found to be mainly due to: (i)

different cutoff values used for categorizing the “wet/dry” and “freeze/non-freeze”
66

 

environments and, (ii) difference between geographical locations and particular climate at
a specific site. The climatic data available for the sites were used to, categorize sites into
four (4) climatic zones according to LTPP definitions for the climatic zones. All the
SPS-l sites were appropriately classified. Figure 4-4 shows the scatter plot between
rainfall and freezing Index (F1) for all sites in SPS-l experiment. It can be observed from
this plot that some of the sites are close to the threshold between wet-no-freeze and wet-
freeze. However, as mentioned above, the cut off defined by the LTPP was used for
location of the sites within different climatic regions.

The as-constructed location of the different sites is shown in Table 4-5. Further, it
can be seen that there are more sites available in wet climate (8 and 6 in “freeze” and
“no-freeze” respectively). There are only four sites in the dry climate (DF and DNF
zones); the two sites present in DF zone are constructed on a coarse subgrade type.
Therefore, the effect of subgrade type can not be determined in DF zone. These
deviations are expected to affect the analysis (the experiment design will become

unbalanced). Consequently, the analysis of the SPS-l experiment design mainly focuses

 

 

 

 

 

 

 

 

 

on the WP and WNF zones.
Table 4-4 Intended SPS-l site factorial [2
W
Subgrade et Dry Total
Type Freeze Non-Freeze Freeze Non-Freeze
T 1A (19)
F‘ OH (39) AL (1) KS (20) NM (35)
me VA (51) 10
‘ M1 (26) LA (22) NE (31) ox (40)
DE (10) FL (12) NV (32) TX (48)
Coarse 8
WI (55) AR (5) MT (30) Az (4)
Total 6 4 4 4 18

 

 

 

 

 

 

 

67

ESE—o E :25“:me 8% 9.56% SE 838m v4» Emmi

Qumruv “3c:— wﬁuooam _a=:=< .w><

 

89 82 88 88 2:. 8m 8
_ 2 r _ _ _ . _
_
20 m 2.0 $8
. O
o u 2
8 an m :88
. mzn
:IIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIII llllllllllllllllllllllll A lllllllll
n O
o u 8.. l8
om .
-
n
O
O 2 “ room
mm . 0 Av
8 8.
O 0 m?» .
-
2 a u 223 :82
-
-
2on
0 :82
2
. 00
—
_
u _o :8:
u
—
n O
u a :82
-

 

 

 

(win) "BJHIBH [enuuv °8AV

68

Table 4-5 SPS-l site factorial — From DataPave 3.0

 

 

 

 

 

 

 

 

 

 

 

 

 

 

Dryb
Subgrade Type c N on- Total

Freeze Freeze Freeze Non-Freeze
IA (19)
OH (39) AL (1) - NM (35)

Fine KS (20) LA (22) 9
MI (26) VA (51) ‘ '
NE_(31)

FL (12)
coarse DE (10) TX (48) NV (32) - 9

AR (5)
WI (55) OK (40) MT (30) AZ (4)

Total 8 6 2 2 18

Note:

P-PP'?’

Design Factors

Wet Regions — Average Annual Rainfall > 20 inches (508 mm)
Dry Regions —— Average Annual Rainfall < 20 inches (508 mm)
Freeze Regions —— Average Annual Freezing Index > 83.3 °C-day (150 °F-day)

Non-Freeze Regions — Average Annual Freezing Index < 83.3 °C-day (150 oF-day)

The design or structural features which are considered to be the main experimental

factors in the SPS-l experiment are:

Each of the above features will be reviewed in this section to identify any

AC Thickness (4 versus 7 inches)
Base Thickness (8, 12 and 16 inches)
Base Type (DGAB, ATB and ATB/DGAB)

Drainage (No or Yes)

deviation from the target values. The asphalt and base layers were targeted for 2 and 3

thickness levels respectively; however, the construction of these target values may

contribute variability in these thicknesses. The amount of variability introduced by the

Construction and how this variability can affect the analysis will be discussed in this

secuon.

69

 

Layer Thickness

The as-constructed asphalt and base thickness were compared with their respective target

thickness. The results of this comparison are given below.

AC Thickness: The SPS-l experiment has two levels of HMA surface thickness — 4-

 

inch (102 mm) and 7-inch (178 mm). The allowable deviation from the target HMA
surface thickness according to guidelines is 6.53 mm. Table 4-6 shows the summary
statistics for each level of asphalt thickness. Among sections with target thickness of 102
mm, the as-constructed thicknesses between all 18 sites has a coefﬁcient of variation
(CoV) of 12.7% with about 43% of the sections within the allowable deviations and 49%
sections having more asphalt thickness than the allowable upper limit. Only 7.5% of the
sections have slightly less asphalt thickness than the allowable lower limit. Similarly, for
the pavement designs which were targeted for 178 mm, the as-constructed asphalt
thickness has a CoV of about 9% with about 78% of the sections meeting the tolerable
limits or having higher asphalt thickness than the upper limit. The frequencies of as-
constructed asphalt thickness are shown in Figure 4-5, whereas Figure 4-6 shows the
scatter of asphalt thickness in different sites within the SPS-l experiment. The overall
low values of CoV for as-constructed asphalt thickness between all sites show that the
asphalt thickness was quite well controllediduring construction, especially for 7-inch (178

mm) target HMA surface thickness.

Que Thickness: The SPS-l experiment has three levels of base thickness— 8-inch (203
mm), 12-inch (305 mm) and l6—inch (406 mm). The allowable deviation from the target
base thickness according to guidelines is 12.7 mm. The summary statistics for as-

constructed base thicknesses at each level are shown in Table 4-6. Among sections with

70

target thickness of 203 mm, the as—constructed thicknesses between all 18 sites has a
coefﬁcient of variation (CoV) of 10.1% with about 65% of the section within the
allowable deviations and 25.3% sections having more base thickness than allowable
higher limit. Only 9.2% of the sections have slightly less thickness than allowable lower
limit. Similarly, the designs which were targeted for 305 mm, the as-constructed
thickness has a CoV of 4.7% with about 79% of the sections meeting the tolerable limits
or either have higher thickness than the higher limit. The designs with targeted 406 mm
of base thickness have a CoV of 4.6% with about 22% of the section having slightly less
thickness than the lower limit. The frequencies and scatter plots of as-constructed base
thicknesses are shown in Figure 4-7. The overall low values of CoV for as-constructed
base thickness for all levels between all sites show that the base thickness was also quite
well controlled during the construction.

The variations between as-constructed and target thickness for asphalt and base
within some sites have shown signiﬁcant difference (see Figure 4-6 and Figure 4-7).
This variation may affect the pavement performance for these sites; therefore, the
deviations between target and actual thickness were taken as covariates in the analysis of

variance.

71

Table 4-6 Summary of comparison between target and as—constructed layer thickness

 

 

 

 

 

Pavement Layer / Count Mean Std C oV Comparison with allowable dev1at1on

Target thickness (inchCS) (%) < Lower limit Wlth tolerab e > U er limit
l1m1t pp

AC Layer

4-inch (102 mm) 106 4.38 0,557 127 < 3.75=7.5% 175-4252134114 >4.25=49.1%

7‘1““ (178 mm) 106 7.12 0.654 9.2 < 6.75=21.7% 6.75-7.25=37.7% ”-25:40“

Base Layer

8-inch (203 mm) 87 8.26 0.84 10.1 <7.50=9.2% 7-50-8-50=65-5% >8-50=35-3%

12-inch (305 mm) 89 11.9 0.56 04.7 <11.5=21.3% 11-5-12-5=66-3% >12-5=12-4%

16-inch (406mm) 36 15.9 0.74 04.6 «55:22.2 15.5-16.5=6l.l% >16.5=16.7%

 

 

 

 

 

 

 

 

 

No. of sections

No. of sections

 

5: V“. "P V?
V‘. ‘l‘ V? "9
m <- <r

AC Thickness (inches)

 

5.5-6

(a) Cumulative frequency for actual AC thickness— 4"target

5.5-6

 

1'2 I: V: 09

\0 tn 1‘ tn

~15 \o' 14 t\'
AC Thickness (inches)

-

 

8.5—

8-8.5 i

(b) Cumulative frequency for actual AC thickness— 7"target

Figure 4-5 Frequency plot for actual AC thickness

72

80%
60%
40%
' 20%

- 40%

100%

Percent sections below

0%

1 00%
80%

60%

20%

Percent sections below

0%

 

 

 

 

 

 

 

 

 

 

 

 

6.0“ . o
”c"? 5.5-1 ,
a)
S o
o 0
$1 5.0- .
CD . o o
m .
cu ' .
.5 45‘ o f o f o
E l 1
4.0 z 1' T , i 4
U .
< 1 3 1 . i. .
3.5- . °
14 51012192 22 26 30 3132 35 394048 5155
Sane
(a) Scatter plot of actual AC Thickness — 4”target
10-
a? 9‘
c1.)
.3: .
o .
.g 85 f }
v .
:2 l l 3 - l l -
E 7' 1 a H . 7
S 6' 0
U .
< 5-
4 T I I I I T I l r I I 1 I f I I I I
l4510121920222630313235394048515

Sane

(b) Scatter plot of actual AC Thickness — 7”target
Figure 4-6 Scatter plot for actual AC thickness by site

73

JNUJLd-b
memo

No. of sections
b
O

15
10
5
0
vitae-0.03%.).
‘9":‘TV1°?W.‘Pa<
sophooooat
Base Thickness (inches).

(a) Cumulative freq. for actual base thickness

 

 

 

      

—8"target
35 IF ¥
30 ] _
.8 25]
H 20 -.
§ 1
¢,_. 15 7 ~
0 L
o' 10l
Z 5 L IL!
0 L .
2 =. L‘- 2. :- =3- :- g;
'7' V? 'T ‘n. 'T V]. 'T —-
9 I: I E S! '2
BaseThickness (inches) .
(c) Cumulative freq. for actual base thickness
—12"target
l4 1 ‘ .
l
12 r _
‘3 10 r l
‘: l .l
a 8 1— l
J.” 6 —
O
c' 4 I» ‘1
z ' ]
2'” L1 1
0 .' - . .
”‘2 "3 ‘r‘. ‘0 V2 ‘\ "7 .r',
E 'T In 'T o T l\ l\'
. l’? 'T "2. 'T ‘n. T -—
1’ ‘2 2 ‘53 E '2
Base Thickness (inches)
(e) Cumulative freq. for actual base thickness
—l6"target

  

Base Thickness (inches) Base Thickness (inches)

Base Thickness (inches)

 

 

 

 

 

 

 

 

 

 

 

 

13< .
124
11-
10]
91 I f _ i
31;}.llll ll,ll}18
1 ‘f t. 1 I 1 I ' i f 'I I I ]
7‘ l i
6‘ r r r v r Y I 1 fr T I I ‘r
14 510121920226303132353940435155
State
(b) Scatter plot of actual base Thickness
—8"target
14-
13 _ T
2?: {311 I };F§[ll 12
t 2 l * 1 l . . 1 s ]
l 1 l i 1 '
11‘ - -
10 Y ‘r ‘7 1 : v I 1 : I v r
14 5 1012 19202263031 32353940435155
State
((1) Scatter plot of actual base Thickness
—l2"target
is] .
t. '. . -
w : ’ _ . . m l 1 . ‘ O o 16

 

 

 

l4 5 l0l219202226303132353940485155

State

(0 Scatter plot of actual base Thickness
—— l6"target

Figure 4-7 Frequency and scatter plots for actual base thickness

74

 

4.3.4 Extent and Occurrence of Distresses

This section of the report presents the availability of the pavement performance
data for all SPS-l sites. The availability of the performance data will be discussed in
terms of extent and occurrence of a particular performance measure. The pavement

performance measures considered in this research include:

a. Fatigue cracking (total area, sq-m)

b. Longitudinal cracking-WP (length, m)
c. Longitudinal cracking-NWP (length, m)
(1. Transverse cracking (length, m)

e. Rut depth (m)

f. Roughness (IRI, m/krn)

It should be noted that various severity levels of the ﬁrst four distresses were
simply added to calculate the total cracking area or length. The change in roughness
(AIRI= IRLatcst-IRIO) was considered in the roughness analysis. The extent (mean distress)
and occurrence (frequency of distress) are presented below for each performance measure

based on data from DataPave (Release 17.0).

Fatigue Cracking
Figure 4-8 shows the occurrence of fatigue cracking in all SPS-l sections by

design and site factors. Based on the latest available data (Release 17.0), about 62% of
the sections have exhibited some level of fatigue cracking whereas about 38% of the
sections have not yet shown any signs of fatigue [see Figure 4-8 (d)]. Similarly, Figure
4-9 presents the extent of fatigue cracking by design and site factors. The distribution of
latest age for all sections is presented in Figure 4-10. It shows that about 10% of the
sections can be considered as young (< 3 years), while the overall average for latest age
of all sections is 6.5 years. Figure 4-11 shows the variation of fatigue cracking within

each site of the $138-1 experiment.

75

E08593 Tmmm .l wcﬁoﬁo osmzmm mo 00:82.80 wé oSmE

wet—8.8 oswzﬂ not. 2388 cc :ouzﬂbma A3

mm. A. mNToS I oonbI
meow D omumND 3.0 B 0-0

Xmm

 

390$ 83 .3 wEono ozwcﬁ co oocotsooo mo “cowxm A8

898,.— “Em

m2 n23 m3
lrl o\oov

U L 0 no
#1 L1: L1: - 11L - LI-
.1 - a\o¢m
- $8
.. $2.
. $8

$8
o\°oc _

 

 

 

panama suonoas JO moored

m
Fm. —1 - .wtl

.lll: litillllcllliu I‘ll
fill] . l l l

 

wExugo oswzﬂ .28 2288 co .8532... CV

:13. 9.2095 9&2?—

m. o o
m w m. n 5 a
> 1 n m 3 a c
o
N
. 8 m
01
. 8 m
w.
8 s
n
a w.
u
8.

 

 

 

Eocene :3er 3 wet—85 ozwzﬂ co 85:38 co 225 AS

£98m .980

I

9Z8

HVDG/ELLV

V

.r o\oov
o\onv
o\oon

$2

$8
$8
-11 -11 1 :3

—L

panama suo grass 30 1U9019d

76

EoEtuaxo 7me I: 880me wcioﬁo =« we 25:25:86 uw< 3+ oSmE

own “moan. .8 95:03 co 8:2:me 3v own 88a. H8 2288 us 55182,; A8

 
 

:s I in E a an a mvu @590 %<
21¢ 95 51m m1m m-
.11. 1.11 o
q - I a
ow
ow
cm
.8:
1 -.|1-‘ l11111-1l11 1. o2

EoEtoaxo Tmmm images osmsﬁ mo anxm aé oﬁmﬁ
880$ swine \3 9.386 uswzﬁ owEo>< A8

282d 8: 3 wﬁxoﬁu oawzﬁ oweo>< 3V
292w..— :wioﬂ

W;
s wmlm

£98... 2%

o a azo an

1. r1 |i1 11.1-11 i. _
q __ .-
8
. . 8
on
8

O

.9.
21

—l.
—t7

30::
c
ocuecwcoooo

LU-bS ‘Sunpeio ensue} aﬁeiaAv
‘ L
_|‘ (IN
' 1
— 91

 

 

uonoas ~10 toqumN

77

w‘bs ‘3”.013'313 311.3118} aﬁeraAv

Fatigue Cracking (sq-m)

 

 

 

 

 

 

 

350«
300‘ ~ as 113
250~
2001
“02 *120

100« i 9‘:

g 1,166 102
50" g a 173 107 113 11:

B - 107 102 ,, 116
1 4 5 101219 20 22 26 30 3132 35 394048 5155

Figure 4-11 Fatigue cracking by site — SPS-l experiment

State

78

 

Longitudinal Cracking- WP
Figure 4-12 shows the occurrence of longitudinal cracking-WP in all SPS-l sections

by design and site factors. Based on the latest available data (Release 17.0), about 46% of
the sections have exhibited some level of longitudinal cracking whereas about 54% of the
sections have not yet shown any signs of cracking [see Figure 4-12 (d)]. Similarly, Figure
4—13 presents the extent of longitudinal-WP cracking by design and site factors. Figure

4-14 shows the variation of longitudinal cracking-WP within each site of the SPS-l

experiment.

Longitudinal Cracking-N WP
Figure 4-15 shows the occurrence of longitudinal cracking-NWP in all SPS-l

sections by design and site factors. Based on the latest available data (Release 17.0), about
68% of the sections have exhibited some level of cracking whereas about 32% of the
sections have not yet shown any signs of cracking [see Figure 4-15(d)]. Similarly, Figure
4- 1 6 presents the extent of longitudinal-NW? cracking by design and site factors. Figure

4-17 shows the variation of longitudinal cracking-NWP within each site of the SPS-l

experiment.

Transverse Cracking

Figure 4-18 shows the occurrence of transverse cracking in all SPS-l sections by
design and site factors. Based on the latest available data (Release 17.0), only 35% of the
sections have exhibited some level of transverse cracking whereas about 65% of the sections
have not yet exhibited any transverse cracking [see Figure 4-18‘ (d)]. Similarly, Figure 4-19

presents the extent of transverse cracking by design and site factors. Figure 4-20 shows the

variation of transverse cracking within each site of the SPS-l experiment.

79

EoEtomxo Tmmm I $5-04 90 022.580 mTv Rama

    

wEv—oﬂo $5-01— ..8 2283 mo :ozzﬂbma A3 @5380 95.01. com 2268 he Accusaoi A3
mN_AI mNToo—I 875.. £65.03
3.10m D om1mND 3.00 90 l L
< m 5. m .75. o
R R m a m u .0
$31. 0
{own om
ow
oo
1 ow
. J LE 9:
.xa xix; es
ov—
ﬂouoﬂ 8; b... mEono $5-04 me 35:38 go 2.me 3v 890m.“ swine .3 met—once @3101— .«o ooeutsouo mo “:25 A3
222i 8% . £88m :93
V
U L LZD “5 L23 m3 .1.
l1 1 .1 1 .11_|1111 1 1 _ 1 LI: a. ILI 1 $0M
_ .. u 1. m;
, :9. m a m m .71. 8 w a w i v
A m Ll lrllr o\cOM
o\.. n 1
_ o m. :2
. 1 $8 M £8
1 u
_ % $9. m , fan?
3
7 W s8 M £3
. as 1 1 1.
1 1 111111 1 11 111111.11 11 1||1ll1 o 1 11111 1 111 11111111111 11.1-1.1
.1 can

uouoas go 1:;qu

80

psxoexo suonoas 50 iuaomd

ﬁestas Sam I as 3 E513 is 2:3

35m
2339; ammzcnaﬁaese n v.

14-3.2--- .
.1

..3_ a 2...
m 1%

 

mean
(9!) (rm-:71

m: x .... 91.. r s .oﬂ
3_ s 3.

2.8
A mo. 8

5mm

 

 

 

 

 

32:5? 2% 1.. .561. .6 Eexm m 3. magma
292d new .3 95-04 omEo>< 3v £28m ewﬁoe .3 55-01— owﬁo>< A3

398 1.— «.380

3085 BE

 

WOO/81V

 
 

 

 

 

 

vm
cm was mm
o
V
w O~W
cm 8m
. 0M9
1W .1...
a.
W em

 

 

O
[\

81

 

EoEtomxu Tmmm l @3210120 835300 21v 859m

wcioﬁo Ab57701. :8 2288 we cocsntﬁo A3 wciuﬂu @3210: :8 meozoommo 3:032“. CV

 
 

 

mN~Al mNﬁiooﬁi OOTWBI 5532-8
WBIOW U OmnmN D mN1o _U 0.. _U I.
< l l a.) Wu K 0
a z I L r.) r .
0.
ON
1 CM
9»
ow
00
on
on

9038 :MEoU .3 mEquo E52104? 83:38 mo Eoexm c3

Eoﬁﬂ 25 3 9.3080 $52.01. me 33:38 go 2.me 3V cw
EoEam €60

mcoﬁum uzm V
:20 “5 :23 1.59

1 s. m 1 1 m m
G 9 7c 8 L V,
e\ooh
LEW .

1 $00

pSXOEJO SUOIIDQS J0 IUQOJOd

 

 

11 :1.\.§

uouoas J0 JaqumN

.xcow
$3
$60
a\omo

$2

°\°
In
I\

P93131213 suonoos }o warned

82

Eustace Sam I an E $2-3 s 3. 2%:

03%
mm .mwvovom mmmm .mOnoNNNONE 99 n

1— . _ _ u . _ _ p _ _ _ - _ _ 1r

f. - _-m «-

ozguw w.
m, T
2—8

-V

l"—

 

E
1
O

1
0
V}

—n
S

S.—

5771“ A...
3 It
8 R
N _
(maxim-3n

com

 

 

 

EQEEExo T mam I 352104 mo «Guamm 21 v 015w: m
$88.: 55% 3 E52- 04 owneo>< A3

E981.— anion—

V
. W;

c
on
9.
co
cm

380$ 8? 3 E5210; mmab>< 9:

£98.“ uzm

u a 1 “:20 .5

{n aﬂemxv

UJ‘

 

 

1111-1111111; 2:

 

O
(‘l

83

‘dMN-o'l aﬁemv

O
O
Lu

“38:83 Tmmm 1| unions 353:8: .8 082580 21v uSmE

mEquu 3.53:9: :8 8233. .8 8:29:35 A3

omAI om1ovl ov16mI OméNU ON1O_U oToD 01D

.Xcmo

   

.1 .
.xa {om

.XL

898: 8: 3 9583 0203:9518 83:38 .8 825 Q:

203mm 25

O 1.. :20 LG m2? m3

o\°o

$9
gem
o\oom
Ax6».
ﬁbn
$8
$2.
$8

 

 

 

panama suonoas JO iuaoxad

wet—23 3831.5: :8 8238 18 xoconcoi E

5.8

< m s m m n

m M M M M m .0
o
ow
9. w
8 m
ow m.
2: m
o~_ m.
o: u
8.

 

€22: cmﬁue 3 95.8.6 033mg: me 85:38 .3 Eme G;

3281* 5:80

N28

1 a :21

w
my

1 n s

l. .V
11111 _- 1&8
1 $3
. econ
$9..
*

1_ $3

P°HOBJQ suouass JO iuamad

84

808898 TQO 1| 0% .3 weioﬂo 385351 omé onE

 

 

 

98m
mm _.m mv ow mm Mn Mm Wm om. cw NW cm a._ me— o__ m w m
1 .1 nu“... 1 m... a .1 1 .1 1 1 o
2. 8... a... .2
m rcN
1cm
.3 m
2... mm. 1% Hm”
18
12.
18
co

 

 

 

8:06:85 Tmmm 1| wagons 8838: .8 Heoexm 21v oSmE

cw a w 3 03.8.. own o> a
298: 85 .3 95.83 omco>8§owﬂo>< 3v £88.: :8 a :Eoﬁo . a c «A v

 

 

 

 

€98."— ozm 398"— 31.60
V
U n. LZQ 115 L2? .15,
1 111.1111 {IrillllLl Tl Fl O
I! 1 V V W
W W Q m m :7... 8 E I. .7
L m GA“ LII «Lib PI l1;
n 2 m H
H m
m 2 n 1.
1“ cm W. 1
1 no
“N w .1111--11-1-111111111111111111 1. 1m

 

1.15.. .1 . 1.

“-1 ‘ﬁuppm 9813ASU31133819AV

85

Rut Depth
Figure 4-21 shows the extent of rut depth in all SPS-l sections by design and site

factors. Based on the latest available data (Release 17.0), about 29% of the sections have
exhibited less than 5 mm of rut depth whereas about 71% of the sections have shown more
than 5 mm rut depth, with 10% of the section showing more than 15 mm of rut depth [see
Figure 4-21 (d)]. Figure 4-22 shows the variation of rut depth within each site of SPS-l
experiment. Figure 4-23 shows the age distribution of all the pavement sections for the

latest rut depth measurement.

Roughness

Figure 4-24 shows the extent of change in IRI (AIRI) for all SPS-l sections by

design and site factors. Based on the latest available data (Release 17.0), about 23% of the
sections have exhibited a negligible change in IRI whereas about 77% of the sections have
shown some level of change in IRI, with 10% of the section showing AIRI more than 0.4
m/km [see Figure 4-21 (d)]. The data is also summarized for the initial IRI (smoothness just
after the construction). Figure 4-25 shows the extent of the initial IRI by design and site 8
factors. It can be seen that about 84% of the sections were built with initial IRI of less than
1.0 m/km and about 16% of the sections with initial IRI more than 1.0 m/km [see Figure
4-25 (d)]. Figure 4-26 shows the variation of roughness within each site of the SPS-l
experiment. Figure 4-27 presents the age distribution of all the sections at the latest

roughness proﬁle measurement.

86

808898 Tmmm I :88 “E 8 88xm _N-v uSmE
£08 2: :8 8233.8 00:3ng A3 :88 S1. .0.: 823318 33:35 A8
E: .530 3.

2A. 212- 21:. :13 in EU 90 2A 2.: 2-: :18 3. E

IIi1_—

 

 

“I

 

 

203$ 02m 3 £30 3: 0mmhu>< 3v 303$ $3.08 3 508 2: emaco>< Aav

903$ “Em

203E cmﬁuﬂ

U m mZO “a "—23 m3

 

(unu) qidaq mu aﬁemxv

 

 

 

N—

o _
om
Om
ov
ow
ow
on

 

suouaas go 3qu

(Wm) Lpdaq mg aﬁenAv

87

“55:0me Tmmm .I #:0885938 ﬁnch HE mo 5:35me om< 84. 239..»

 

own 282 Le 2263. he coznebma 3v

2-9. 950 WmD m-m! m-D

$2

 

:6

own 382 .8 3283. go hocoscoi A3

a4.

Eoﬁtaxu .Tmmm II 8% 3 59% EM mmé 2sz

nmﬁmvovommmmm—momommmom

85m

m_o_mv_

p

 

5

a.
E...

 

ma

f

“M
a

a 2:.

WI!“

”1

t
:—

 

#3

3

i

m2...

a L.

2...

.o

 

@533 %<
Wm

(“1111) mdaa m‘d

m-m

 

suopoas JO 1:5qu

88

 

EoEEono Tmmm I22 mo anxm vmé oSmE

P: E omcmcu Sm 2288 Mo :ozzebma A3

_ A I de I wduod I c.o-v.o D v.0-N.o D Nduo D o- D

ﬁxﬁm

   

Boaoﬂ 8i 3 3: E omega owEo>< 3V

$98“. “Em

U m LZQ m0

m2; m3

 

cod

o_.o

omd

Omd

ovd

(“DI/VJ) 131V same/xv

S: E ems—Eu .8 30:03. go hogs—uni 3v
:35 .22
.0

8'0'9'0
V'O'Z'O

1'8'0

.0
.0
Z

9'0"?7

suopoas J0 .19qu

fiil} ‘ xiix‘ ‘ ‘ co
E98.“ :wmmou B 5: E uwcmzu umﬁu>< A8

Eouuam 5300

9:
‘ zx

(“DI/“1) an =3wAv

89

 

208593 3% Lace swim m3 2:5

5: 35:: L8 228% go 8:2:me A3 5: EEE Low 2283. .«o 5532... CV

EVE .05:

I--I.__.
I

222a :wiut an 5: EEE owEo>< A8

L

$28; 87. 3 5: EEE owEo>< 3v

ESoE ozm Eouunm summon

V

L2G no "~23
.LIII II I_I I I; I Ir I I?! I! no m
w

00
I “—1? IL!

E

2!

.II

m

PI I Ll

w

(unvm) onﬁ aﬁmsAv

 

#__u “,1- .

I.

J—
o
\O
suopaas }o JaqumN

v4 3
9o

, no
M3

, g

l
D
-—'

(um/m) 01m aﬁeJaAv

90

“5&5ng Tmmm I: “582388 3,05%on mo cousﬂbﬂu um< nmé oSwE

own «mos. .5.“ 228% mo 22:5ng 3v

2-0- aéLu Wmﬂ m-mD m-D
o\oNv
.XEN

 

XE

o\oa .XKL

own 323 F8 828%.? 35:35 AS

 

 

 

EuEEono Twmm I 86 E mmuﬁwnom cmé 235

2; E 5: E25 3V
28m
mmSwvovonmmﬁﬁﬁﬂmmﬁﬁ~_o_ w v—

 

 

iva j;

 

I}!

.000

.26

row.—

 

5

(MW) r211 IBRIUI

 

Eng»?
.9 3. 3. E m-
I Irll II 0
Ii l 2
ON N
- 8 m
9. m
- on m.
8 m
I on m
I Ox 5
I om
oo.
2; 3 v: E omcﬁo 3
88m
mm __n w... ow mm Mm mm _.m om ow ~.~ ow m. ~._ o._ m my _
I m I 8:. WAS
gun» a. an in:
2... t u Ind
m .. E. m
2:. m 8. ”ﬂ” 8... .0..
8... .2
no...

 

 

 

(unI/w) 1211 H! 03mm

91

4.3.5 Dynamic Load Response Data (DLR) — Flexible Pavements

This section of the report summarizes the data availability for the instrumented
ﬂexible pavement sections in OH (39). According to Ohio University report [6], eight
series of controlled truck tests had been completed on these instrumented pavement
sections as shown in Table 4-7.

Each series of tests followed a similar pattern with regards to how the tests were
setup and conducted. The general steps followed during each test are discussed in
reference [6].

Only series 11 data and part of series IV data are available in DataPave (Release
17.0). Also data pertaining to instrumented SPS-8 sections in Ohio are not available in
the database. The testing setup details have been obtained from DLR_TEST_MATRIX
table. The locations of strain gauges and LVDTs data were obtained from
DLR_STRAIN_CONFIG_AC and DLR_LVDT+;CONFIG tables. The depth at which -
strain gauges were installed is not available in the DataPave; therefore this data was
obtained from Ohio University report [6]. The peak strain, deﬂection and pressure data
were extracted from DLR_STRAIN_TRACE_SUM_AC,
DLR_LVDT_TRACE_SUM_AC and DLR_PRESSURE_TRACE_SUM_AC tables.
Only data collected from these instrumented sections in 1996 and 1997 are available in
DataPave. The speciﬁcs of the tests during series II (in 1996) are listed in Table 4-8 and
Table 4-9. The test dates for which strain data are available in DataPave are shaded in
grey. Table 4-10 details the series IV test sequence, which is available in the Release 17.0

version of DataPave.

92

T 4-7

Test
Series

. Parameters
Section

Monitored

Test

Dates Truck

No. Passes
Load Speed

Tires

I CNRC 144 l X X

II 6 X X

30-60/sec

Single
Tandem 50

FWD drops/sec.
Dynaﬂect 20

Load I.D

 

 

 

 

 

 

Table 4-9 Series H Truck Parameters - ODOT Tandem-Axle Dum Truck
Rear Axle Load I
, (kiPS) _
Date Nominal Load Nominal Speed Load LD Run No.
(RIPS) (mph)
Lead Rear
8/2/96 32 16.62 16.23 30,40,50 A 1-17

 

 

 

 

 

 

Table 4-10 Series IV Truck Parameters — ODOT Single-Axle Dum Truck

 

    

Seamus]

93

Nominal

Speed (mph)

Load I.D

 

It was also observed that not all the runs conducted during each test series for a speciﬁc
date and sections are available in the strain data (see details in Appendix A3 of reference
[1]). Furthermore, strain data is not available for all gauges or all speed levels. For
example, in section 39-102 data recorded for only 3 strain gauges are available in the
database, whereas, the instrumentation plan for this section shows that there are 6 strain
gauges located under the asphalt layer. Appendix A3 of reference [1] shows the average
peak strain values of the data for different offset categories.

Data summaries were also prepared for surface deﬂection data (from LVDT) and
pressure data (from pressure cells) within each section, and are attached in Appendix A3

of reference [1].

Discrepancies in Dynamic Load Response Data

The data availability for dynamic load response for the 'SPS-l experiment in the
currentversion of DataPave has highlighted several discrepancies. These deﬁciencies
can seriously affect the usefulness of this data for any type of analysis; some of these
shortcomings are highlighted here for future improvements:

0 Keeping in consideration the amount of data collected for these instrumented
sections; only limited data from series II and series IV are currently available
(DataPave Release 17.0).

o The direction of strain gauges (Longitudinal or transverse) is not available in
DataPave. These had to be obtained from the Ohio University report [6].

0 Similarly, the depth of strain gauges from the surface is also currently missing
from the database. 1

o In order to validate the dynamic load response mechanistically, the material
properties for pavement layers have to be calculated at the time of testing. To
facilitate this objective, the time of testing and temperature should be included as
a part of dynamic load response data. Also it would have been useful to have data
from FWD testing conducted at the locations at the strain gauges and pressure
cells.

94

4.4 DATA AVAILABILITY IN SPS-8 EXPERIMENT -- FLEXIBLE
PAVEMENTS

This section of the report summarizes the data availability for ﬂexible pavement
sections in the SPS-8 experiment. The availability of all relevant data elements is
summarized in Table 4-11. The table shows that availability of the main factors is high,
while that of the exogenous factors is somewhat lower. In particular, the availability of
trafﬁc data is low; however, the impact of trafﬁc data may be insigniﬁcant for the SPS-8
experiment if all the sites have very limited trafﬁc. The availability of the relevant data

elements in SPS-8 experiment is discussed in the following sections of the report.

4.4.1 General Site Information

Each site has unique characteristics, which can be mainly explained by the
particular climatic and soil conditions at a particular location. The SPS-8 experiment
mainly focuses on pavement performance based on the environmental aspects of sites in
combination with different subgrade types. The particular site information can be further- -

divided into construction, climate and trafﬁc.

Construction Issues

The construction reports prepared by the supervisory consultants for each site
were reviewed to identify the deviations/problems during the construction of each site.
These deviations might be helpful in explaining the unusual trends in performances

(premature failures) at a particular site.

The summary of deviations has been prepared for all 15 sites in the SPS-8

experiment and is given in Appendix C of reference [1].

95

Table 4-11 Summary of SPS-8 data element availability —Flexible pavements

 

 

 

 

 

 

 

Data Category Data Type Data A‘Ljuabmty’
0
Construction Reports 93
Climatic data
Virtual Weather Station
Annual Temperature 93
Annual Precipitation 93
Automatic Weather Station
Site Information Monthly Temperature 47
Monthly Precipitation 47
T raﬂic data
Traffic Open date 93
Estimated ESALs 60
Monitored ESALs 33
Axle Load Spectrum 33
Asphalt Layer
Core Examination 80
Bulk Speciﬁc Gravity 75
Max Speciﬁc Gravity 78
Asphalt Content 78
Asphalt Resilient Modulus l9
Penetration 69
. Viscosity 65
Material Data Asphalt Speciﬁc Gravity 69
Aggregate Gradation 81
Fine Aggregate Particle Shape 21
Layer Thickness 100
Unbound Base Gradation 78
Subgrade
Subgrade Gradation 73
Atterberg‘Limits 84
Subgrade Modulus 44
Layer details
Type 100
Representative thicknesses 100
Constructed thicknesses 100
Pavement Structure Shoulder information
Width 93
Thickness 93
FWD data
Deﬂections 100
Temperature at Testing l00
Monitoring“ Backcalculated Moduli 13
Manual Distresses data 100
Longitudinal Proﬁle (IRI) 100
Transverse Prcjile (Rut Depth) 100

 

 

Note: .. Data is said to be available for a section even if it is available for one survey.

96

 

Climatic Data

As explained before for the SPS—l experiment, the average annual rainfall and average
annual fre- ezing index were used to classify each site into four climatic regions. The
classiﬁcati on deﬁnitions for each zone were taken from the LTPP DataPave. The
summary of the climatic data from the VWS for all sites in SPS-8 is given in Table 4-19.

Only climatic data for CA (6) is not available in Release 17.0 of DataPave.

T raffic data

The SPS-8 experiment design stipulates that trafﬁc volume in the study lane be at
least 100 vehicles per day but not more than 10,000 ESAL per year. Therefore, it is
important to check the trafﬁc not exceeding the threshold speciﬁed for this experiment.
Theitrafﬁc data is only available for 8 out of 15 sites from estimate and monitoring
modules 0 f DataPave (Release 17.0). No trafﬁc data is available for AR (5), CA (6), MO

(29). NJ (34), NM (35), NC (37) and WI (55).

4.4.2 Material Data

The material pr0perties of all the layers in a pavement system play a very
Signiﬁcant role in its future performance. The SPS-8 Experiment was designed to study
the SPeCiﬁc effects of a range of environments on the pavement performance; therefore
the material properties which are susceptible to climatic changes need to be investigated.
In this GXpeI‘i ment the subgrade type was a factor (ﬁne or coarse), while the asphalt mix
and base material properties were assumed to be uniform across all states. The subgrade
material properties were investigated. The summary of soil gradation and Atterberg limit

information required for classiﬁcation is given in Table 4-13.

97

Table 4- 1 2 Summa of Environmental data of the sections in SPS-8

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

Avg. Avg.
State Climatic AATP' AIPDZ WDPY3 Days Days AAT‘ Fl FT
Zone (mm) (days) (days) Above below (°C) (deg days) (cycles)
- 32°C 0°C
050800 WNF 1374 34 133 64 52 17 46 48
080800 UP 372 7 - 95 31 162 10 326 142
280800 WNF 1427 37 145 52 65 16 57 60
290800 WF 1079 27 144 37 105 13 167 92
29A800 WF 945 22 137 29 1 12 12 334 84
300800 DF 371 4 132 4 198 6 574 163
340800 WF 1071 27 119 8 68 13 127 56
350800 DNF 346 5 92 83 99 15 9 100
360800 WF 891 17 193 5 130 9 437 87
370800 WNF 1342 33 151 36 46 17 14 47
390800 WF 972 24 153 10 130 10 374 96
460800 DF 423 8 96 25 175 7 978 107
480800 WNF 1015 24 131 99 19 20 10 18
48A800 WNF 846 22 100 94 35 19 21 34
490800 DF 473 7 ' 118 ’8 198 7 498 170
530800 WF 510 7 137 30 91 11 169 73
53A800 WF 386 3 135 33 88 11 163 71
$0800 WF 814 17 . 151 4% 175 6 1015 96

 

 

Note: -
l-Average Annual Total Precipitation (mm), 2-Average Intense Precipitation Days in a year, 4-Wet Days per Year,
4-Average Annual Temperature

98

A53 .mogov— 6% N131. 88:08 gnu; Got L8 mzocowcmn on 2 EE mod :9: 5:: Etoﬁﬁgm— SE 22: 9.32 >20 3.68 5%
3.35 m m m D 95 83: 53: 69:53 3303 33 55:5 £5 £8 03335 .8 .EAE a. $2A E850 5:0on

  

oz<m ozé- so- So 8.9.9: SN
639% 3am é :8

”802

38

D_ RZIW 22m.
3. 22$

99

In addition, two critical aspects of soil behavior were further investigated from the available
soil data: eXpansion of clayey soils in dry zones and frost susceptibility in freeze zones. The

active so i 18 were identiﬁed by using the following criteria [7]:

- Expansive Soils— colloidal content >15% and PI>18

- Frost Susceptible Soils— silt, coarse clay having > 15% material ﬁner than

0.02 mm. ,
By using the above criteria, the subgrade soils in States 29 (Missouri), 46 (South

Dakota, section 0804) and 49 (Utah, section 0804) were classiﬁed as active (expansive)
soils, while sections in States 5 (Arkansas), 29 (Missouri), 39 (Ohio, section 0804) and 46
(South Dakota) were identiﬁed as having subgrade soils with frost heave potential.
4.4.3 Design versus. Actual Construction Review

According to the original experiment design as discussed in chapter 2, 12 sites were
essential required with two different structural designs. These sites were selected based on
the geographical location so that they may be located in different climatic regions. However,
due to site speciﬁc climatic data the region-identiﬁed at the design stage may be different.
Similarly, the target layer thickness may have variability due to construction. The speciﬁc

as-construc ted site conditions are discussed in the section below.

Constructio ’1 Issues
The construction guidelines for SPS-8 sections were discussed in chapter 2. The

COHStructiotl deviations for each site were taken from the construction reports and are

summarized in Appendix C of reference [1].

100

Site Factors

The SPS-8 ﬂexible experiment design required that two different structural designs
should be repeated in at least 12 sites. However, the actual data on the site factors (climate
and subgrade) showed that there are 15 sites in the SPS-8 ﬂexible experiment and currently
these are distributed according to Table 3-4. There are 7 sites in WF, and 3 sites each for

WNF, DF and DNF zones, respectively. Almost half of the sites were constructed on coarse

subgrade, and the others were built on ﬁne subgrade soil.

Design Factors

The design or structural features which are considered to' be the main experimental

factors in SPS-8 ﬂexible pavement experiment include:

0 AC Thickness [4-inch (102 mm) versus 7-inches (178 mm)]
- Granular Base Thickness [(8-inch (203 mm) versus 12-inch (305 mm)]

The summary of the as-constructedand target thicknesses for all ﬂexible pavements

is given in Table 4-‘14.

4.4.4 Extent and Occurrence of Distress

The age of the section is a very important factor in the SPS-8 experiment, as a higher
age Of a particular section will translate in higher environment related distresses. Figure 4-28
ShOWS the latest age for all ﬂexible pavement sections in SPS-8. Further age distribution of
ﬂexible pav ements among the SPS-8 sections is shown in Figure 4-29. The age data for
BPS-8 SectiC) 115 shows that most of these sections are aged below seven years and are in the
early Stage 011 the performance curve.

The distress data for the SPS-8 sections was obtained from the ﬁles

MON_DIS_AC_REV (cracking and non-load related distresses data),

M0N_T_PR0F_INDEX_SECT10N (rutting data) and MON_PROFILE_MASTER

101

(roughness). Figure 4-30 and Figure 4-31 show the occurrence and distribution of distresses
in the SPS-8 Experiment ﬂexible pavements. The available distress data in Data Pavel

(Release 17) has only shown ﬁve types of distresses in all SPS-8 ﬂexible pavements.

Figure 4-31 (a) shows the distribution of rutting in SPS-8 ﬂexible pavement sections.
It can be observed that only 9% of the sections have shown more than 5 mm of rutting,
where as in the majority of the sections (60%) rutting ranges from 3 to 5 mm. A low amount
of rutting is expected in the SPS-8 pavements since load is the major cause of rutting in

ﬂexible pavements.

Figure 4-31 (b) shows the distribution of roughness data (IRI) based on its
magnitude. The data suggests that the majority of sections did not exhibit high levels of

roughness, with only 9% of the population with IRI greater than 2 m/km.

102

Egble 4-14 Construction details of the ﬂexible pavement sections in SPS-8

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

State SHRP_ID sugife AC GB 031 332 T8 Tiget Target
5 0803 F 3.8 7.3 4 8
5 0804 F 7.2 12.7 7 12
6 A805 C 4.2 8.2 4 8
6 A806 C 6.6 12.2 7 12
28 0805 C 4 9 4 8
28 0806 F 7 12 7 12
29 0801 F 4.9 7.8 4 8
29 0802 F 7.5 11.5 7 12
29 A801 F 4.3 8.3 4 8
29 A802 F 6.9 12.3 7 12
30 0805 C 4.5 7.1 4 8
30 0806 C 6.9 11.8 7 12
34 0801 C 3.5 7.8 4 8
34 0802 C 6.8 11.6 7 12
35 0801 F 4.4 9.7 4 8
35 0802 F 7.3 12.6 7 12
36 0801 C 4.9 8.4 168 4 8
36 0802 C 7.6 10 156 7 12
37 0801 C 4 8.7 4 8
37 0802 C 7 11.5 7 12
39 0803 F 3.9 7.9 36 4 8
39 0804 F 6.6 11.9 30 7 12
46 0803 F 4.8 8 4 8
46 0804 F 7.2 12 7 12
48 0801 F 4 8.5 10 4 8
48 0802 F 5.5 10.7 10 7 12
49 0803 C 4.9 7.8 41.2 4 8
__ 49 0804 C 6.9 12 41.2 7 12
_ 53 0801 F 3.7 8 38.4 4 8
_¥ 53 0802 C 6.8 11.7 38.4 7 12
__ 55 0805 C 4.4 8 4 8
55 0806 C 7 12 7 12

 

 

 

 

 

 

 

 

 

 

 

NOte: ll-Granular subbase, 2-SS represents subgrade layer, the thickness in this column is the ﬁll

103

 

 

 

 

 

 

1

wowoumm
mowoumm
monoumm
Swoumm
vowc$v
mowo6¢
Nowoév
Swoév
vowoév
mcwoév
vowoém
mowouom
Nownzh
53-?“
Nowoém
awoém
Nowo-m m
Swoﬁ m
mowoém
Swoém
wowoém
mcwoém
mow/cram
_ow<-mm
Nowoumm
Swoém
cowoém
mow0-wN
ccw<é
mcw<uo
vowoum
mowo-m

Section ID

 

 

12

 

10

 

6
4
2-
()J

m

6%

25%
44%

 

 

 

 

Figure 4-28 Age of the ﬂexible pavements in SPS-8

   

 

 

 

 

@

.

_

_

.

m
J
4

«5:830 62

D<2 [212-5 US-7U7-9I>9

5-7 7-9

A8608“)

26

(b) Distribution of age

(21) Frequency of age

Figure 4-29 Age distribution in the SPS-8 sites — ﬂexible pavements

104

  

69%

I Alligator Cracking D No distress

(a) Fatigue cracking

 

I Long. Cracking (NW?) D No distress

(c) Longitudinal cracking-NWP

53%

38%

  

62%

I Long. Cracking (WP) D No distress

(b) Longitudinal cracking-WP
22%

  

78%
I Transverse Cracking D No distress

((1) Transverse cracking

22%

  

78%
I Ravelling D No distress

(e) Raveling -

Figure 4-30 Distribution of distresses in SPS-8 ﬂexible pavements sections

 

I-lIl-3U3-SUS- I-lIl-ZDZ-3U34

(a) Average Rut (mm) (b) Average IRI (m/km)

Figure 4-31 Distribution of IRI and Rutting in SPS-8 ﬂexible pavements

105

CHAPTER 5 - METHODOLOGY FOR ANALYSIS

5.1 INTRODUCTION

The purpose of this chapter is to provide a summary of analysis methods that were
used to perform this research. Some of the previous studies analyzed LTPP data (GPS
and SPS experiments) based on engineering criteria (“basic” statistics) and subjective
judgment [1-4]. For example, the engineering criteria may include the rate of growth,
severity levels and impact of distress on the functionality of the pavement. Several
statistical methods were employed for establishing performance criteria to study the
effect of design and construction features on pavement performance in this research. The
statistical methods range from trend plotting to complex multivariate analysis.

This research focuses on evaluating the effects of speciﬁc design and
constructions features on the response and performance of ﬂexible pavements (SPS-l
Experiment). The selection of statistical methods was based on the speciﬁc objectives of
this research and the extent and occurrence of performance data. These methods, as well
as the concept of Performance Index (PI) developed and employed in the analysis, are

explained within this chapter.

5.2 PERFORMANCE INDICATORS

The performance of a pavement is an accumulation of damage over time. All
pavement sections within each SPS-l site were monitored over time; however, the
monitoring of these sections is staggered with age (i.e., the distress data were collected at
different times for individual sections), and the performance measures (cracking, rutting
and roughness) have shown a variable trend with time. Therefore, it was felt necessary to

develop a measure that can quantify the overall performance of a pavement section over

106

time. Figure 5-1 through Figure 5-4 show various performance curves (cracking and
rutting distresses) for twelve test sections within two sites [Alabama, AL (1) and Iowa,
IA (19)] of the SPS-1 experiment. These ﬁgures show the measurement variability of
distresses with time. The following discussion presents various options that were
considered to transform the time series data of a section into a single performance
indicator. The Options considered are listed below:

Maximum distress at the latest age/survey.

Area under the performance curve.

Area under the performance curve normalized to the latest age.
Performance Index.

Maximum distress at the latest age is one of the options used for time series data
analysis. This performance indicator only considers the maximum distress that was
recorded for the test section in its monitored lifetime. Also, this performance indicator
will not capture the performance trend over time and will provide only a snapshot of
performance at a given time. In addition, the measurement variability over time is not
taken into account.

Area under the curve represents the actual pavement performance for a distress;
larger area indicates poorer performance. The area under the curve can be calculated
with the trapezoidal rule by using Equation (5-1).

n+hn

5-1
2 ( )

Area = ; (’z‘+1 " ti)

The shortcoming of “area under the curve” is that this indicator cannot
discriminate the performance of two sections having the same area but with different

times for distress occurrence. For example, the performance curves in Figure 5-5 and

107

Figure 5-6 may have similar “area under the curve” but the curve in Figure 5-6 shows
better performance than that in Figure 5-5.

Area under the curve normalized to the latest age can be another alternative
which can eliminate the discrepancy of using “area” alone (as mentioned above). This
indicator can also be calculated based on the trapezoidal rule and can be represented
mathematically by Equation (5-2).

2 (t. — t. ) ——y" + y‘“
Area i 1+1 ' 2
L = i L

age age

 

(5-2)

Where; "Lage’, is the latest age used to normalize the “area”.

This indicator distributes the performance of a section (area) evenly over all
years. However, performance curves can exhibit highly variable trends with time (see
Figure 5-1) and may have gaps in the data for some years. Therefore, an alternative
indicator was selected, where the performance is weighted with age.

The Performance Index (PI) is deﬁned as:

Zn "1'

P1 = 1 (5-3)

Z‘z‘
I

Where:
ti = the age at distress measurement year i

y,- = distress measured at year i(for example alligator cracking in, sq—m, rut depth in mm
and IRI in m/km)
Note that only the ages at which distress measurements were taken are included in

the calculation of PI. Equation (5-3) can be further simpliﬁed to the form of a series as

shown by Equation (5-4).

108

 

Alllgata' Craklng (sqm)

 

 

 

Age (years)
+0101 +0102 +0103 +0104 +0105 +0106
+0107 —0108 +0109 +0110 +0111 +0112

Figure 5-1 Fatigue cracking with age— AL (1)

H
on
O

 

H H

A 0"

O O
[Tnﬁl‘rr

Long. Crad<s WP (m)

H H

N 48 01 (D O N

O O O O O O
:‘rrﬁ rrrr—rrr rl’I‘r—V‘le V-F'I'YT—I'I'IT'T—TW r

l

 

Age (years)

+0101 +0102 +0103 +0104 +0105 +0106
+0107 —0108 —0109 +0110 +0111 +0112

Figure 5-2 Longitudinal cracking—WP with age -— AL (1)

109

Rut, mm

Trans. Craddng (m)

W -5 U! m \l on
O O O O O O

N
C

 

 

10'

 

 

 

 

Age (years)

+0101 +0102 +0103 +0104 +0105 +0106
-+-0107 --0108 ---0109 +0110 +0111 +0112

Figure 5-3 Transverse cracking with age — IA (19)

 

 

 

 

Age
+0101 +0102 +0103 ~‘~~ 0104 +0105 +0106
+0107 +0108 +0109 +0110 +0111 +0112

Figure 5-4 Rutting with age — IA (19)

110

Cracking Area

    

h----—-

 

c-F
I--|
0+
M
1-0-
m
c-t-
4:
H
U]

Years

Figure 5-5 Poor Performance

Cracking Area

       

p--—---

 

‘
r

 

 

t5 t6 t7 t8 t9

Years

Figure 5-6 Good Performance

111

V

«I‘

PI-yl't1 +y2't2 +y3't3 + ...... +391,-

_th 29 25' Zr.-

 

(5-4)

It can be seen from Equation (5-4) that higher weights will be given to the
performance measured at the later ages (as ti+1 > ti and 2 ti is constant for a given

pavement section). This makes the performance index more applicable to the SPS-l
experiment, which stipulates that no maintenance or rehabilitation action should be taken
during the life of the pavement sections. The following hypothetical example illustrates
the comparisons between various performance indicators discussed above.

Figure 5-7 shows the performance curves for ﬁve different pavement sections.
The best and the worst performing sections are to be identiﬁed from the time series data.
Three of the performance indicators were calculated for all ﬁve sections and the results
are summarized in Table 5-1. It is clear from the results that section D is best performing

because the distress remains at the same level over the years and all indicators are
capturing this well. The second best section according to “Area” and “Area/Lage” is

section B; however according to “P1” section A is second best.

By visual comparison of the performance curves for sections A & B, it can be said
that section B will deteriorate at a faster rate compared to section A, given the
performance history of the sections (see Figure 5-7). As higher weights are given for later
years in the calculation of the Performance Index (PI), it is expected that this indicator
will be more suitable to capture the present and relative ﬁiture performance of each
section. Therefore, P1 was selected from among various performance indicators for this

study.

112

The performance indices (PIS) were calculated for each section and for the different
performance measures such as cracking, rutting and roughness. This was calculated by
summing the product of distress and age for all available surveys and dividing it by the
sum of ages for available surveys, as shown by Equation (5-3). All analyses (overall and
site level) were performed using PIs for test sections.

Although PI seems to be the best Option among all the performance indicators
considered, it has some inherent limitations. These limitations are mainly because PI is
dependent on the number and timing of the distress surveys. For two pavement sections
of the same age, with one monitored each year and the other monitored on alternate
years, the PI for the former section will be slightly lower. For the same sections if the
monitoring was not performed at a regular time interval, the section with more surveys
towards the later age will have a slightly inﬂated PI. However, this limitation may not
have considerable impact in the case of the SPS-1 experiment as all the pavement
sections were monitored with a regular time interval of 1 to 2 years. In SPS-1 and SPS-8
experiments the pavement sections at different sites have different ages. Among
pavement sections with different ages and similar performance, the PI of younger
sections will be lower than that of older sections. To address this issue, the age of test
sections was considered as a covariate in all statistical analyses of PI. This will adjust the
P15 according to the age of pavement sections. The statistical methods used in the study

are brieﬂy explained next.

113

 

    

 

 

 

 

 

 

 

 

 

i w 7
14"---------" --------- ,x’a --------- L ---------
12E . _ .. ’ .. )4”’_. . E ----. i .- _
.. .t ‘ ’ I’ ‘ 3 3
glot - ...... ' --
a i x ;
'c 3 ------------------------------------ r ---------
s i .
a -:_'L': .. ‘ ..........
g 6 r i

. g ............ -

t
2 E -------------- 4
O 1

o 2 4 6 8 1o 12

Age (years)
"9" Section A + Section B - i- ' Section C + Section D + Section E

Figure 5-7 Comparing different performance curves — An Example

Table 5-1 Calculated performance indicators
Performance Sections
Indicator A B C D E
Area 39.75 i; 67.5 19 53
Area/1,6128 4.0 i; 8.4 2.1 5.9
PI ' 2.; 5.9 10.9 2.0 7.5

 

 

 

 

 

 

114

 

5.3 OVERALL STATISTICAL ANALYSIS METHODS

The type of statistical analysis depends on the objectives and nature of data
collected for a particular research question or hypothesis. Generally, the vast majority of

data likely to be encountered in research fall into the following seven designs[5]:

One-sample data

Two sample data
K-sample data

Paired data
Randomized block data
Regression data
Categorical data

There are four basic questions needed to classify a set of data into one of the

5 even models listed above; these are:

Are the observations (variables) quantitative or qualitative?
Are the units similar or dissimilar for different variables?
How many treatment levels are present in the data?

Are the observations (variables) dependent or independent?

The dependent (target) variables (response and performance measures) in the
SPS-1 experiment design are quantitative, whereas the independent (explanatory or
attributes) variables are qualitative. Figure 5-8 is a ﬂowchart that summarizes the model-
identiﬁcation process. It can be observed from this model identiﬁcation process that the
data from SPS-l experiment falls into the category of randomized block design or K-
sampled model. Figure 5-9 shows a summary of research tasks and respective
hypotheses to achieve speciﬁc objectives within each task. Depending on the extent and
availability of data speciﬁc statistical methods were employed to investigate the research

questions. Figure 5-10 summarizes the data analysis methodology for overall analysis.

115

More than 2

 

Start

 

 

 

 

 

Are the data
qualitative or
quantitative?

 

 

Qualitative

Categorical
data

 

 

 

Quantitative

 

Are the units
similar or
dissimilar?

 

 

Dissimilar
J Regression
I data

 

 

 

Similar

 

How many

 

 

treatment levels
are involved?

 

 

O
ne One-sample
data

 

 

 

Two

 

1

 

 

Dependent

 

 

Are the samples
dependent or
independent?

 

 

Are the samples

 

 

dependent or

 

independent?

 

 

 

 

Randomized
block data

Independent

 

 

 

Dependent Paired
data

Independent

 

 

K-sample
data

Two-sample
data

Figure 5-8 Summary of the model identiﬁcation process [5]

116

 

20% :50}.
.m> .0.... ”maniac?
3.85: .n> $2.495.

3.2.2: .m> E. o
.0:... a .a “059.2: 25...
22.85.55?
3532.. 855.83..
52.? 9.53200

3:38:20.
Esau—gm o
232 .1922. E... 8: 5
3532: 3588.2. 55:
E... 0532: oo.5....8.on
ﬁlm a :oqun
:o_.oa.o.:<n.£m:o_5_o.
50.23.. .95 0.0... w—

mxmm. £0.88. .80 88m Wm mama

 

 

A»... :9

85302 855.85..
.m> 53330.5 0

2:262
355.83.. .m> Ema o
.05.. a 5 2.0 .3532:
3:88.. c5. 8558.2.
52.5. 9:32.50

>335.an o

3:23... o

:32 x833: c5. 2...

.s 355.85: :5. 3:83.
525.3 5:32.20.
59.25... b... 0.0.: «—

 

 

 

ason.“ .50..—

.oE: a .a 20 .8532:
3:88. 9... 8588.2.
25...; 9.103250

m_m>_u:< w:_.uo:_m:w o
meoﬂiqﬁoo 3.5...— o
magniﬁcent

oi .. 5.23 mcmﬁou 52:5
.8 2.88:: 8:83.

.5: 3558.2. u... u.<

 

 

.0:... a .u 2.: £0.3qu
8:83. .5... 3558.2.
58.? actuEmcoU

nuutm 5:222:— o
Euotm £22 0
$8.08 2;
c5 :o...u=..m:8\:m_muu
2.2....»
.8 50.8.6 3:88.
:5 3558.2. u... o.< .
”£82.83.

 

 

 

 

 

 

 

5.352;: “£358.”: 65050:»: $522“ :a..u>Ov
8.55.8.5.— .n> 3.5.5.8.: 3555...»..— .m> 8.330”— _u>ow_ 9.5 .254 3.83.92
65.. 35‘. .051— _o>oi_ 635 651. $51. 65:.
0:03... Z 02m .0382 “Em 0:03.02 ”Em €03.02 8%

 

.8 e. S

OUCNELOVHDL
.m> OUCNCEOMCUL

 

     

   

5 .m> A:
85:..85m
.m> 88538

3

DUCNFCOmHDn—

 

S

omcoamom

 

  

29:35. «:5

  

117

 

Overall Analysis

 

 

 

 

 

 

 

7

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

Pavement Performances Pavement Performance
over time Index
!
ANOVA Survival Pavements Pavements
Repeated Analysis with zero with > zero
measures Time to failure distress distress
distributions
Kaplan—Meier (Classiﬁcation W [AN OVA
Survival Compare 0 Treating main
Functions pavement sections factors as
with and without ﬁxed effects
distress o Treating site
Cox Proportional 0 Contingezncy factors as
Hazard Model , Tables-x test random
of effects-
independence blocking
- LDA variable
0 Logistic
Regression \ J

 

 

 

k J

r Non-Parametric Test

0 KS Test for checking normality

0 Sign Test difference between two
distributions

0 Wilcoxon Test

\ J

 

 

Figure 5-10 Methodology for overall analysis

118

 

 

 

As shown in the ﬂowchart (Figure 5-10) for overall analysis, two types of methods were
used for overall statistical analysis. One is based on the magnitude of the performance,
i.e. comparison of mean performances between the levels of various factors. The other is
based on the frequency of occurrence of distresses, i.e. probability of occurrence or non-
occurrence. The AN OVA (one-way and multivariate) method belongs to the ﬁrst type.
The Linear Discriminant Analysis (LDA) and the Binary Logistic Regression (BLR)
belong to the second type. These methods in the context of this research are discussed

below.

5.3.1 Analysis of Variance (AN OVA)

The ANO VA is a tool that allows for better understanding of how the independent
variables (categorical) inﬂuence the dependent variable (continuous). Using the General
Linear Model (GLM) univariate procedure, various hypotheses can be tested about the
mean of a single dependent variable when cases are classified into groups based on one
or more factors (independent variables). For example, the effects of different base types
or asphalt layer thickness (factors as independent variables) on the amount of cracking
(dependent variable) are candidates for such an analysis. Moreover, some of these
independent variables may be considered to be having a fixed or a random effect on the
dependant variable. Also, any other continuous variables (independent) for which the
dependent variable is to be adjusted can be included in the model as a covariate. Both
balanced and unbalanced models can be tested by AN OVA. A design is considered as a
balanced design if each cell in the model contains the same number of cases.

ANOVA can be performed by considering one factor at a time, or by considering

more than one factor at a time. ANOVA is “one-way” when the effect of a single factor

119

on a dependant variable is studied, while it is “multivariate” when the effect of more than
one factor on a dependant variable is studied. Also, multivariate ANOVA is more
efﬁcient as it adjusts for the effects of various factors at a time. Moreover, interaction
effects, if any, between various factors can be studied by multivariate ANOVA.

ANO VA Diagnostics (Assumptions): It is not necessary, nor is it usually possible (when
dealing with ﬁled data) that an ANOV A model ﬁts the data perfectly. ANOVA models
are reasonably robust against certain types of departure from the model, such as the error
terms not being exactly normally distributed [6]. The major purpose for the examination
of the apprOpriateness of the model is therefore to detect serious departures from the
conditions assumed by the model. To apply ANOVA models, the observations must be
independent random samples from normal population with equal variances. The residuals
can be used to check these assumptions in order to have conﬁdence on the observed
signiﬁcance levels. The residual plots can be helpful in diagnosing the following
departures from AN OVA model [6]:

Non-constancy of error variance
Non-independence of error terms

Outliers

Omission of important explanatory variables
Non-normality of error terms

Generally, two common departures from ANOVA model— non—constancy of the
error term and non-normality of the distribution of the error terms, are found in the data.
The following frequently recommended remedial measures are found in the literature [6-
8]:

- Often, non-constancy of the error variance is accompanied by non-normality of
the error term. A standard remedial measure here is to transform the response

variable (dependent variable). Two approaches are normally considered to ﬁnd
the appropriate transformation to accomplish normality and constant variance—

120

some simple guidelines (e. g., log, natural log or square root etc.,) and the Box-
Cox procedure (an iterative procedure to ﬁnd the best power for transformation).
0 If the error terms are normally distributed but the variance of the error term is not
constant, a standard remedial measure is to use weighted least squares.
0 When there are major departures from the AN OVA model and even

transformations are not successful in stabilizing the error variance and error

normality, a non-parametric test for the equality of the factor level means may be

used instead of the standard F -test.

All the assumptions of the ANOVA models used in this research were checked
and appropriate remedial measures as discussed above were adopted where ever
necessary. For example, ANOVA model used to identify structural factors affecting the
change in IRI has residual scatter as shown in Figure 5-11. The residual scatter against
predicted value shows a fan shaped structure, indicating violation of the constant variance
of error terms. Similarly, as mentioned above this violation of constant variance of error
terms is accompanied by the non-normality of the error terms as shown by Figure 5-12
and Figure 5-13. A natural log transformation of the dependent variable (change in IRI)
was adopted as a remedial measure. The same ANOVA model was executed on the
transformed variable and again the residual plots were examined. Figure 5-14 shows the
residual plot from the model. It can be observed that no serious violation of the constant
variance of the error terms is detected and residuals seem to be randomly distributed
along the predicted values. Figure 5-15 and Figure 5-16 show the diagnostics for
normality of the error terms. It can be seen from these plots that residuals are also
normally distributed. Therefore, the conclusions from ANOVA model based on the
transformed change in IRI were only considered in the research. The same methodology

was employed for all the performance and response measures in ANOVA models. It

should be noted that the number of data points may be reduced when log transformation

121

is adopted; as the sections having zero distress will be eliminated from the analysis. The
consequence of this elimination is ﬁirther discussed in Chapter 6.

Statistical Significance: Statistical signiﬁcance of an effect of a factor implies that there
exists a signiﬁcant mean difference between the performances (in this study) of any two
levels within the factor. For example, a statistically signiﬁcant effect of HMA surface
thickness on fatigue cracking implies that there is a signiﬁcant (statistical) mean
difference between fatigue cracking on sections with 4-inch (102 mm) HMA thickness
and sections with 7-inch (178 mm) HMA thickness. Moreover, in simple terms, a
statistical signiﬁcance indicates that the effect is not a happenstance. Regardless of the
type of analysis the p—value identiﬁes the likelihood that a particular outcome occurs by
chance. The smaller the p value, the greater the likelihood that the ﬁndings are valid. It
is generally accepted that if the p value is less than 0.05, the result is considered
statistically signiﬁcant. Thus, when there is less than a l in 20 probability that a certain
outcome occurred by chance, then that result is considered statistically signiﬁcant.
Another corrrmon convention is that when a signiﬁcance level falls between 0.05 and
0.10, the result is considered “marginally signiﬁcant”. When the signiﬁcance level falls
far below 0.05 (e.g. 0.001 or 0.0001 etc.), one will have greater conﬁdence on the
research’s ﬁndings. A signiﬁcance level of 0.05 was used in all the statistical analyses
performed in this research. However, it is important to conﬁrm the practical or
operational difference between the means of various levels of a factor, if a factor has a
statistically signiﬁcant effect. An attempt was thus made in this study to gauge the
practical or operational signiﬁcance of statistically signiﬁcant differences in the analysis.

The operational signiﬁcance adopted for various performance measures is discussed next.

122

Residual for Delta IRI

 

 

 

 

 

 

I o
o
o
0
0° 0
-0.S-r- 0
o
.11»
0.0 0.5 1.0

Predicted Value for Delta_IRl

Figure 5-11 ANOVA diagnostics-Residual plot without transformation

 

60— Mean = O
__ Std = 0.235
" N=212
so—
40...
5‘
g: /\
3 30—
l-
Ia.
20— —

 

 

 

 

 

 

 

 

 

 

 

 

l T
-0.5 0.0 0.5 1.0 1.5

Residual for Delta_IRl

Figure 5-12 ANOVA diagnostics-Residual frequency without transformation

123

Normal Q-Q Plot of Residual for Delta__IRI

 

 

 

 

0.75
O
o
0.50- 0 0° 0
0 o o
2
a
> 0.25ﬂ
T:
E
‘6 000-
z .
1:
0
H
8 025»
n. o
N
ea
.050“ 8
O
-o.7s : 1 t t
4.0 -o.5 0.0 0.5 1.0 1.5
Observed Value

Figure 5-13 AN OVA diagnostics-Residual normality without transformation

 

 

 

 

 

 

1.o-~ o
o 0 o
o o
0- ' o O o o
o o
E 51 ° 6’ 800 ° ° 0° °
... o 00 o o o o
5 Q3 °°° a? 00 99:00 0°
9 o Qboo o (p A 0 Q
o o o

a: o
A 0°00 9° 0 8 o
l- 0 O O O 00 O O O
02 O O O . o O O 0
Eli-0.5" o o 0 o
.3 0 o o
.g Q 0
m

-m- o

-l.5-- I,

.2.0 -15 4.0 .05 0.0 0.5

Predicted Value for Ln_DeltaIRl

Figure 5-14 ANOVA diagnostics-Residual plot with log (natural) transformation

124

 

30— Mean = 0
' Std = 0.32

N= 163

 

 

25— 7‘\

i

Frequency
if

i

 

5- \

\

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

-1 .o 41.5 0.0 0.5 1.0
Residual for Ln_DeltaIRI

Figure 5-15 ANOVA diagnostics-Residual frequency with log (natural) transformation

Normal Q-Q Plot of Residual for Ln_DeltaIRI

 

.0
2‘

Expected Normal Value
$

.5
T

 

 

 

 

Zr .5 -r.o cs 0.0 0.5 1.0
Observed Value

Figure 5-16 ANOVA diagnostics-Residual normality with log (natural) transformation

125

Practical Signiﬁcance

The statistical signiﬁcance of the difference between the marginal means for
various levels of design and site factors needs to be judged from a practical (operational)
point-of-view. This practical signiﬁcance is dependent on the magnitude of the mean
difference of levels for a particular factor and will vary for each performance measure.
For example, if the means for alligator cracking are signiﬁcantly (statistically) different
for pavement sections constructed on DGAB and on ATB, one should check whether this
difference has any practical or operational meaning from an engineering point-of-view.
The practical signiﬁcance therefore depends on the subjective judgment of actual
pavement performance observed in the ﬁeld.

To determine reasonable levels of practical signiﬁcance for different distress
types (fatigue cracking, rut depth, transverse cracking and roughness), the performance
curves developed based on engineering judgment of an expert panels were used. These
curves for various distress types were developed under a previous study [1, 9]. The
criteria for fatigue cracking, rut depth, roughness and transverse cracking performances
are shown in Figures 5-17, 5-19, 5-21 and 5-23, respectively.

As mentioned before, the AN OVA was conducted on PI for all performance measures
except roughness. For roughness the change in IRI (Latest IRI- Initial IRI) was used as
dependent variable in ANOVA. Because the marginal means from ANOVA are in terms
of PI, the performance curves from the expert panel were converted to P1, assuming 1
year monitoring interval. These curves, in terms of PI, for fatigue cracking, rut depth and

transverse cracking are shown in Figures 5-18, 5-20 and 5-22, respectively.

126

5...... S. .o E .8 8.8.... 85.8.8.0. ow-
$.85 02

 

 

 

wagon... 83.8 .o .n. .8 3.2.... 8.88.8.9. w.-m 0.5m...

o.

 

1.1.4..“1 .11...\li.ll !

   

m 0.5%....

 

u.Sm.0.:.-coZ|wl -
0.580.... |¢l _

 

 

.28.. .3

m m N. c

 

M

 

m.

 

.1! on

 

1111.!
.\..\..,\.

 

 

 

 

 

_

28m.o.:.-:oZlmrrﬁ . ii .

2.3.2:. lol
.11.. 11 1.

 

 

 

 

 

(umMIbCIIQIJOH

(%) Summa anﬁpva so Id

T. 5...... .3. .8 8.2.... 8.58.8.8. m . -m 0.5m...

o. m w h

3.85 om<

o

m

 

 

 

 

 

 

 

 

1.1

0.8885252 lml .,
0.882... l0!

.

 

..
vrlLim. ._.J.LJ_Lmr

 

 

 

o.

m

o

.

N

S. 3.88... 2&8... .8 8.3.... 8.88.8.9. :8 0.3m...
_ .23.. a...

w

 

1!! .11!

_
W. ,
_
_

.
1 A 1.

 

 

 

i—l

. 2822:.-eoZIml . .
2.3.2:.Iou

       

_
.
_
.
A
_
.
.
.
.

 

 

 

o.

(um) Wham

(%) Summa 9113925

127

E 0382880888. .8 8......0 8.88.885. mN-m 08m...

 

 

889309..
c. m w n o n v m N . o
. . 1.1.! 4 . _ I J o
.
08.828-802Iml . . . .. . ; md

    

0.8.0.5 Ion

  

 

(uni/m) m1

 

\ f. a-.- i m.N

”J

 

 

 

 

 

 

 

..lz!! -LI -I-1 m

88.8... 8.888.. .8 E .8 8.2.8 8.88.88... ma 8%.... E 98.8.”. 8.888.. .8 8.8.8. 8.88.8.8. .m-m 8:»...

$890 ow<

 

 

 

(w) smavds Homo
(w) qﬁm x3213
(w) Sumds m0

 

(m) mﬁm X32030 Id

 

 

 

 

 

 

 

 

 

 

 

128

It can be seen from these curves that the slopes of the individual performance curves vary
with age. For example, the $10pe of the IRI curve is the same up to year 5 and later can be
separated into two parts. From these two slopes, change in IRI per year can be calculated
for the ﬁrst ﬁve years and for the next ﬁve years (see Figure 5-23). The weighted average
of these slopes was used to calculate the change in IRI per year.

Furthermore, the above described curves deﬁne the boundaries between good and
poorly performing pavements for interstate and non-interstate highways, respectively. For
the SPS—l experiment, it was estimated that 80% of the designs correspOnded to the
interstate highway class, while the remaining 20% were non-interstate, based on the
asphalt layer thicknesses. Therefore, the slope (change per year) was further weighted
for the proportions of the pavement class within the SPS-1 experiment. Table 5-2 shows
the threshold values for practical or Operational signiﬁcance for the various distress types.

Another important aspect of AN OVA models is the sample size and statistical
power to detect the means difference between the levels of a factor. This issue of sample

size and statistical power is discussed next in the context of this research.

Statistical Power and Sample Size (Replication)

Unfortunately, this aspect of the statistical analysis is ignored in most of the
studies involving data from the actual ﬁeld, mainly because of the economic and practical
considerations. Sample size is usually determined from statistical considerations,

resource or budget considerations or both. Generally, the larger the sample size, the
greater will be the ability to detect any difference in the response due to different

treatments.

129

Table 5-2 Operational si iﬁcant differences for various performance measures

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

Performance measure Weigh ted slope Remarks
per year
Fatigue cracking (%) 0 20 This will translate into 1.0 sq-m of area
' per year.
Rut depth (m) 0 80 The operational signiﬁcant difference for
' rut depth is 0.8 mm per year.
Transverse 3 50 This will translate into 75 m of crack
cracking (m) ' spacing per year.
The weighted slope was calculated based
on 5000 ft/mile failure criterion used in
NCHRP l-37A. The failure criterion is
Longitudinal 4 50 thus 144 m for a SPS-1 test section. The
Cracking (m) ' operational value is based on the SIOpe of
the performance curve between 0 and 10
years, assuming zero cracking up to 5
years and failure at 20 years.
Roughness The change in IRI was calculated based on
AIRI(“‘/km)'Flex‘ble 0'13 initial IRI and latest IRI
True State of Nature
H0 is true H1 is true
Correct Type 11
g: Accept H0 Decision Error
2%
8
Q
5
. Type 1 Correct
Re} ect H0 Error Decision

 

 

 

Figure 5-24 Type I and Type II Errors

130

 

Thus a key step in experimental design is to assess the power of the statistical test to be
used in the analysis of the data or conversely, the precision of the estimates to be
produced by the analysis, as a function of the sample size. Eventually, a trade-off must
be made between the increase in power and precision resulting from the higher sample

size and added cost or time required to complete the experiment.

Flaming of sample size can be approached in terms of (1) controlling the risk of
making Type I and Type II errors, (2) controlling the widths of the desired conﬁdence
intervals or (3) combination of both approaches. Type I error (0:) occurs if the null
hypothesis (H) is rejected when H, is actually true. On the other hand, Type II error ([3)
occurs if one fails to reject Ho when the alternate hypothesis (H 1) is actually true. Figure
5-24 shows the four possible “decision-state of nature” combinations. Generally,
researchers do not directly discuss the probability of Type 11 error ([3); rather, they use
power, which is equal to l-B. Since [3 is the probability of the event of incorrectly
accepting (or failing to reject) the null hypothesis when it is not true then the probability
of the complement of that event (correctly rejecting the null hypothesis when it is not
true) is l-B, which is called the power of statistical testing. Thus “the power of a
hypothesis test is the probability of correctly rejecting the null hypothesis when the null
hypothesis is not true”.

For any statistical analysis, one would like to have a test that gives low Type I
error rates and high power, but one cannot have both with a limited sample size due to
ﬁnancial constraints. Therefore, a power of 70% to 80% is considered reasonable for
most of the scientiﬁc studies. The power for any statistical analysis can be calculated
given the sample size, an estimate of the standard deviation and expected mean

131

difference. Conversely, for an assumed power of the hypothesis, the sample size can be

calculated. When a single population is involved with unknown variance, the power of

hypothesis test can be calculated using Equation (5-5).

lac—#1“;

POW8r=Pf T> ta/z df ——-—-——-— (5‘5)
’ 5

Where;

Ho: p=u° versus H]: “attic: u,
df= n-l

n: sample size

Equation (5-5) can also be modiﬁed to Equation (5-6) to calculate the sample size

(n) for a given power.

( 2 2

. -. )s

2 a/2 l—,62 (5-6)
(yo-”#1)

As mentioned above another approach based on the desired range of conﬁdence

 

interval is also used for calculating the sample size. Equation (5-7) can be used to
calculate sample size for a single population with unknown variance.
2‘2 32

/ 2 d

= _‘Z__’f_ (5-7)
2
E

Where;
E = margin of error, which is equal to half width of conﬁdence interval
df= n-l
n = sample size
When two population means are involved in the statistical analysis with unknown

variances, Equation (5-8) can be used to calculate the power of test when sample sizes are

unequal, whereas Equation (5-9) is for equal sample sizes.

132

 

 

 

 

 

Power-l—ﬂ=Pr t>t — '60 -51I (5-8)
- a/2,df 1 l
\/MSE [— + —]
x "1 "2 1
Power = Pr t > t — 60 - 51' (5-9)
and!" 2MSE .
72

Where;

H0: 1.11-1.12 =60 versus H]: pl-pz 1:50: 51
‘ df= n1+n2-1 for unequal sample size

df= 2 (n-l) for equal sample size

n: sample size

Equation (5-9) can also be modiﬁed to Equation (5-10) to calculate the sample

size (n) for a given power. -

 

 

2 2
(ta/2 —t1—,6) (25p )
n 2 2 (5-10)
(50 " 51)
Where:
2 2
Sp = pooled variance = df 1x31 +df ZXSZ
df1+df2

Note: Pooled variance is only calculated if both populations gave equal variance

Similar to the case of single population, the sample size can also be estimated

using the margin of error approach, as shown in Equation (5-11).

2

2
2
n=’ (a/2,df) (SP) ’ (5-11)

E2

 

133

0 Factors aﬂecting Power

From the above mentioned equations for calculating statistical power and sample
size, the following factors should be considered for achieving reasonable statistical power
in any statistical analysis:

0 A smaller value of or will decrease the power. This means that one will
always trade-off Type I against Type 11 error.

0 If the alternative mean is further away from the null mean the difference
ape — p, l) increases, then the power increases. Note that if the means are

close, then it is less likely to reject the null hypothesis; in fact, failing to
reject the null hypothesis may not have serious consequence if the
difference is not practically signiﬁcant.

0 Power decreases with increasing variance (oz). The variance can be
controlled by selecting homogeneous experimental units (pavement
sections) and by obtaining replicate measures for each pavement section
and using the average of several replicates for an individual section.

The power calculations are somewhat more involved for one-way AN OVA and
multi-factorial ANOVA models. However, the interpretation of power concepts remains
essentially the same as explained above for statistical analyses involving one and two
sample populations. The methodology for power calculation using these ANOVA
models can be found in references [6, 10]. Statistical Computer Software such as SASTM

bTM

and Minita can be used to calculate the sample size (replicates) for a given power of

an experiment design involving many factors or a single factor with many levels.

Figure 5-25 is an example of power plots for a one-way ANOVA, in which a
single factor (designs) with 24 different levels was considered. These 24 treatments

represent all possible combinations of the designs in the SPS-1 experiment.

134

 

L

Power

IIVIITIWIII

T I

.o
N
r

 

1 [11:41 1 J 'lllALl

10 100

Mean differences

(a) Error Variance = 25

 

 

 

 

 

 

Power
.0 .O
A o

.0
oo

/'

 

//

 

 

 

p—

10 100

Mean differences

(b) Error Variance = 100

 

Power
.0 .0 P .0
N -h 0‘ 00 '—

FTTTTT WT!— Y‘ﬁ’TT'TT’]“T—T.r—I "YT—['7‘

O
p—n

1000

 

Llilll' A '1‘il'll

10 100
Mean differences

(c) Error Variance = 625

 

+2
+3
+4

 

+6

L____"""7

 

 

1000

Figure 5-25 Effect of replication and variance on statistical power—One-way AN OVA

135

The power curves represent three levels of error variance (see Figure 5-25 a, b and c),
which can be represented by the mean square error (MSE) -— an estimate of error
variance in the one-way model. The power curves were plotted for different levels of
replicates in the model (n=2, 3 ...7). It can be seen from these plots that a lower mean
difference to attain the same statistical power can be detected either for a low variance or
for higher number of replicates. Also, higher mean differences for the same number of
replicates are associated with higher statistical power. For example, to attain a power of
80% in this design with 0:10, about 4 replicates will be required for each of the 24
design to detect a mean difference of about 35 between any two of the designs. If the
replicates are increased to 7 for each design, a mean difference of about 25 can be
observed with a power of 80% [see Figure 5-25 (b)]. In order to detect a lower mean
difference (less than 15) for 80% power, the error variance has to be reduced and number
of replicate need to be at least 7. A few suggestions to reduce the variance between the
sections were mentioned above.

The SPS—1 experiment design involves multiple factors; therefore to simulate this
experiment, power curves were generated for a design having six factors with two levels
each. These curves for three levels of error variance ((52 = 25, 100 and 625) are shown in
Figure 5-26. The standard deviation (0) for fatigue cracking within SPS-l experiment is
about 10 sq-m. To detect a means difference of at least 5 sq-m of fatigue cracking
(minimum level of practical signiﬁcance) between the levels of factors at 80% power
assuming 02 = 100, a replication of at least 2 is required within each cell of the
experiment [see Figure 5-26 (b)]. However, a single replication of each treatment will

yield a power of less than 20% to detect the same mean difference. Furthermore, if o is

136

greater than 10 sq-m of fatigue cracking, a single replicate will have a very limited power
to detect any mean difference less than 70 sq-m and even 2 replicates of each treatment
will not be able to detect a mean difference of 10 sq-m of cracking at 80% power.

It should be noted that the above discussion of sample size (replication) and
power is only valid for a balanced design, i.e. no missing value for any of the cells in the
experiment. However, the actual scenario with real data may reﬂect a more complicated
situation; for example Table 5-3 shows the real replication of the test sections for fatigue
cracking at this point in time. It can be observed that the experiment design is quite
unbalanced because of the many missing values for treatments. The implication of this
unbalance on multi-factor AN OVA models may seriously hamper the power of the
statistical analysis. Therefore, simple statistical analyses such as comparison of means
between two populations or one-way AN OVA (where more than two means need to be
compared) are warranted at this point in time. This suggestion strongly applies to
cracking performances since a considerable number of sections have not yet shown any
distress (see Chapter 4 for details on data availability).

The simple analysis will have higher statistical power as the sample size will
increase when some factors are ignored. However, these methods may not adjust the
means for factors other than the ones considered in the analysis. Figure 5-27 shows the
power curves for comparison of means between two populations. Each curve in the
ﬁgure represents a detectable mean difference. To detect a mean difference of 5 sq-m for
6:10 sq-m at 80% power, a sample size of at least 60 each for both populations will be
required [see Figure 5-27 (b)]. However, a sample size of about 20 is required to detect

the same mean difference at the same power for o=5 sq-m [see Figure 5-27 (a)].

137

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

F
0.8 E r._l
,
IL; 06 E —o—2
3 —~—3
o E
; +5
0.2 : —-—6
‘ —+—7
0 '41111111 L L1 4.11 L P1 1111 1 1 441L111
0.1 1 10 100 1000
Meandifferences
(a) Error Variance = 25
l
E m
0.8 ”If
E —o—l
,.
‘- 0.6 ““2
g E +3
Q4 0.4 E‘ +4
E —-—5
0.2 *5
E .4 —o—-7
O l 1 1 4 LLLLLL 1 1 1 1 1111 1 L 1 L1 1-11 1 14+L114
0.1 1 10 100 1000
Meandifferences
(b) Error Variance = 100
l E r
0.8 E- .
.. -+—l'
‘- 06 l --.—22
g l- —~—3E
f' +53
0.2 E +5}
L --r—7i
0 L 1.. 1 .
0.1 l 10 100 1000

Mean differences

(0) Error Variance = 625

Figure 5—26 Effect of replication and variance on statistical power—Multi-Factorial
ANOVA

138

Table 5-3 Actual replication of sections within SPS-1 experiment— Fatigue cracking

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

D . B T Base HMA WF WNF DF DNF Row
ramage ase ype Thickness Thickness F C F C F C F C Total
8.. 4" 1 2 1 2 1 1 8
7n 4
DGAB 4" 1 1 1 1 6
12" " 1 1 1 I 1 1
7 1 2 1 1 1 1 7
8.. 4" 3 1 1 1 1 1 8
. 7" 4
No ATB 4" 1 1 1 1 6
12'! " 1 2 1 1 1
7 3 1 1 1 1 7
8.. 4" 2 1 1 1 1 6
7" 5
ATB/DGAB 4" 1 1 1 1 1 5
12" n 1 2 1 1
7 3 1 1 1 6
8.. 4" 1 1 1 1 1 1 6
7" 1 l 1 1 4
4" 5
PATB/DGAB 12" .. 1 1 1 1 1
7 3 1 1 1 1 7
4H 5
16" 7" 1 l 1 1 l 4
Yes .. 1 1 1 1
8.. 4 1 2 1 1 5
7" 3 1 1 1 6
4H 6
ATB/PATB 12" .. 3 1 1 1
7 1 1 1 1 4
4H 4
16" n 2 1 l
7 1 1 1 1 1 5
Comm-[om 32 26 16 20 o 23 4 12 133

 

139

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

' 5 1
[+10-
1... 151
O l ‘7 1 1 1; g L L L
l 10 100 1000
Sarnle Size (n)
(a) PoOled variance = 25
l
0 8 1 1/ / / / ' 5
1., 0.6 5 +15
5 : +20
‘1‘ 0.4 : +25
1
0.2 3 +30
: v
0 1 1 1_ 1 1 1 L l 1 ‘ 1 1 L L1 ' L 1 ,1” 1..- 7 ,1 44444,-
1 10 100 1000
Samle Size (n)
(b) Pooled variance = 100
1 r
.1
0.8 E
E +5
5; 0.6 E +10
E +
o. 15
0.4
E —+-—20
0.2 . +25
E +301
0 . 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1n_ - - - 1 1 14 1
1 10 100 1000
Sarnle Size (n)

(c)‘Pooled variance = 100

Figure 5-27 Effect of mean difference and variance on statistical power—Test means for
two populations

140

Based on the statistical power considerations and data availability for the SPS-1

experiment design, a simple univariate may be more suitable for detecting a lower mean

difference (for various performance measures) between the levels of experimental

factors.

Discussion on Analysis of Variance (ANO VA)

The analysis of variance (AN OVA) is a powerful method; however under certain

conditions (limited available data) the application of this method has some restrictions.

These issues are brieﬂy summarized below:

The un-balanced data makes it difﬁcult to meet the equal variance
assumption. Therefore, an appropriate transformation of the response
variable may be adopted to address this issue.

In case of fractional factorial design, the higher order interactions

(between more than two factors) cannot be studied [11].

Replication within each cell of the experiment design plays an essential
role in determining power of hypothesis testing. The power of a
hypothesis test is the probability of correctly rejecting the null hypotheses
when the null hypothesis is not true. Lower number of replications within
experiment design will reduce the power of detecting a mean difference
between levels of a factor for a given variance.

Time series AN OVA with repeated measures seems to be an appropriate
choice of analysis for this type of experiment where each section is
monitored over time. However, this type of analysis requires a more
balanced data i.e., all pavement sections should be monitored at the same
interval and up to an age long enough (about 15 years) for capturing long

term pavement performance.

141

Comparison of Means (One-way ANO VA)

As mentioned before, the SPS-1 experiment was designed as fractional factorials.
Further, since the number of sites constructed in each zone-subgrade combination is not
the same and not all the sections are exhibiting cracking distress, the experiment is
unbalanced. Thus only two-way interactions may be reliable for the experiment. In
addition, transformation of the response variables becomes an essential choice to fulﬁll
the requirements (assumptions) of AN OVA. When natural logarithmic transformation is
applied to response variables, due to the nature of the transforming function (natural
logarithm of zero or a negative value is not deﬁned), only data pertaining to those test
sections that have distressed (i.e. non-zero positive data) were considered in ANOVA.
Hence, AN OVA results are based only on distressed sections. 1

In the SPS-1 experiment, each SHRP ID represents a unique design and thus there
are 24 designs in the experiment. The performance of the designs with respect to each
other was evaluated using the deviation from mean performance, which is the standard
deviate. The designs were evaluated based on selected distresses, considering one distress
at a time. The standard deviate was calculated for each of the twelve designs within each
site. Table 5-4 shows a sample calculation for alligator cracking in the AL (1) site, which
is calculated by using the following equation.

PI of a given design - Average PI for the given site)

Standard Deviate = ( . , , ,
(Standard Dev1at1on of P15 for the given Slte)

 

As this measure was calculated for each section, considering one site at a time, it

indicates the relative standing of the section compared to other sections.

142

Table 5-4 Calculation of standard deviate for alligator cracking - Alabama (1)

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

Section ID Performance Average Standard Standard deviate
Index deV1ation
0101 23.08 -0.70
0102 90.69 1.36
0103 25.43 -0.62
0104 3.83 -1.28
0105 95.90 1.52
0106 16.12 -0.91
0107 21.97 4593 3282 -0.73
0108 70.45 0.75
0109 75.47 0.90
0110 52.74 0.21
0111 65.33 0.59
0112 10.22 -1.09

 

143

 

This transformation thus helps nullify the variation in performance (due to site
conditions) among sites, as the sections are weighed with respect to companion sections
in each site. The standard deviate will show the relative comparison of various designs
for a speciﬁc performance measure. This value can be interpreted in the following three
possible ways:

0 Lower value indicates better performance than the mean

0 Zero value indicates the mean performance

0 Higher value indicates worse performance than the mean.

The standard deviate for a particular performance can also be used to compare the
effects of design factors, and for this one-way AN OVA was performed on the standard
deviates of the sections. The analyses were performed on data from all sections and also
on subsets of data stratiﬁed by different subgrade types, climates and combinations of
these. This helps identify the effects of design factors under different site conditions. The
standard deviate values of each design were averaged from the various sites to study the
overall as well as the interaction effects of climatic zones and subgrade type.

To consider all available sections, the test sections were categorized as “cracked”
and “non-cracked” for frequency based methods (LDA and BLR). These analyses
methods (discussed below) will help in identifying the signiﬁcant factors that

discriminate between the two categories.

144

5.3.2 Extent of distress

The effect of the key experimental factors on performance, through the
relationship between the magnitude and relative occurrence of the observed distresses,
can be observed from the data. Simple bivariate plots between the percentage of test
sections that have exceeded various levels of distress for the key performance measures,
categorized by experimental design and site factors were plotted to display and explore
the data. Note that the effect of climatic zone will be only shown for the wet regions

because of the limited number of sites (4 sites) in the dry regions.

5.3.3 Linear Discriminant Analysis

Linear Discriminant Analysis (LDA) allows for distinguishing between two or
more groups of data. This is done by identifying variables that are signiﬁcant in
classifying the data into various groups. The procedure for predicting membership is to
initially analyze pertinent variables where group membership is already known. The
details of theoretical background of LDA is available in literature [7, 12, 13]. For
example, groups of observations can include one group of pavements with cracks and the
second group with no cracks. The method allows for determining which variables

discriminate between cracked and non-cracked pavements.

5.3.4 Binary Logistic Regression

Binary Logistic Regression (BLR) is used often in the case where the outcome
variable is discrete (dichotomous). The difference between logistic and linear regression
is reflected both in the choice of a parametric model and in the assumptions. This

method is based on the maximum likelihood method for determining the parameters of

145

 

ar

511'

of.

interest. The details of theoretical background of BLR are available in literature [14]. The
interpretation of effects for various levels of the categorical variables (independent) is
very convenient in terms of the odds ratio when this type of model is used. Logistic
regression models are also very useful for discrimination analysis (of various groups)

when categorical variables are used as independent variables.

5.4 SITE-LEVEL ANALYSIS METHODS

The site level analysis evaluates the performance of each section based on
comparisons between the similar designs with in a site (state). It is assumed that within
each site, climatic conditions, subgrade soil type and trafﬁc volume are identical for all
test sections. Thus the main advantage of this analysis is that comparisons are made
among those sections that were subjected to similar loading and environmental
conditions. Furthermore, construction methods, material sources and surveys are also
assumed to be identical within each site. All site-level analyses were conducted using the
Performance Indices (PIS) of the sections for various performance measures. The
difference in performance is assessed based on average values. The details of analysis
are discussed below.

Comparisons by Design Factors

The site-level analysis consisted of series of comparisons, each focusing on the
effect of a particular design/construction factor, for SPS-1 experiment. Such comparisons
are not possible for SPS-8 sections because of the limited number of sections in the
experiment. For the site level analysis, each performance measure was analyzed in terms

of its performance index (Pl).

146

Comparisons were done at two levels—A and B. In level-A analyses, all designs (0101
through 0112, or 0113 through 0124) at a given site are compared such that only one
factor is held common within the sections of each group. For example, in level-A
analysis, the effects of HMA thickness (4” vs. 7”) were studied, within a site, by ignoring
base type & thickness, and drainage.

In level-B analyses, most of the factors are ‘controlled’ for comparisons. In other
words, individual sections within a given site are paired such that all but one design
parameter are the same. This parameter is the factor being studied. Comparing a given
pair of sections will allow for determining the effect of the particular design factor with
the Mghest possible level of constraint (level-B). In this case, there are four factors being
studied, so the highest possible number of constraints is three. For example, comparing
sections 111 and 112 (SPS-1) allows for determining the effect of base thickness (8”ATB
versus 12”ATB), while comparing sections 216 and 220 (SPSZ) allows for determining
the effect of base type (DGAB versus LCB). Table 5-5 and show possible comparisons
within a given site in the SPS-1 experiment.

The relative effects of levels within each design factor were studied based on the
ratio of mean performance of the sections corresponding to a level over the mean
performance of all levels of the factor. A sample calculation of relative performance is
presented in Table 4-6. In the table, the comparison of relative performance indicates that
pavement sections with 7” HMA surface thickness are performing better than those with
4” HMA surface thickness, since the relative performance is lower for sections with 7”

HMA surface thickness (0.8 versus 1.2).

147

lo

1111

DE

ef.

th:

1111

W

 

For factors with two levels (such as HMA surface thickness, PCC thickness, and
drainage), the relative performance of each level ranges from 0 to 2, a value of 1
indicating no effect of the factor (i.e., the amount of distress (performance) corresponding
to the two levels of the factor is the same). A value less than 1 indicates better
performance compared to mean performance of sections corresponding to both the levels
of a factor. Consequently, a value higher than 1 indicates the worse performance. The
best possible performance translates to 0, and the worst possible performance translates
to 2. For cases where there is no distress, each level of a given factor will have relative
performance of 1 indicating no difference in performance.

For factors with more than two levels, similar logic can be extended. For the
effect of base type, the relative performance of each base type ranges from 0 to 5, since
there are ﬁve base types in SPS-l experiment. In the case of the SPS-2 experiment, for
the effect of base type, the relative performance of each base type ranges from 0 to 3,
since there are three under comparison. A value of 1 for all base types indicates that the
amount of distress is the same for all base types. Values close to l for all the base types
being compared indicate that there is no signiﬁcant effect of the base type. A higher value
indicates more distress (worse performance) for a particular base type. For SPS-l, the
worst possible performance translates to 5 (all other base types would show 0, indicating
no distress), and the best possible performance translates to 0 (no distress).

The relative performance for various levels of the main factors was calculated for
all the sites in SPS-lexperiment, and for each performance measure. The concept of
relative performance can be utilized across the sites without considering trafﬁc variability

because it is calculated at site-level.

148

 

 

 

 

 

 

 

 

 

 

38E: dz E. .2 2 a $on5 dz Go. .2 8:
m: dz a: .2 2 3 m5. dz $2 .2 8:
3.8 dz $: .2 2 c 3.8 dz E: .2 a: m
«8525 $05.25 .
33 vi mctocﬂ Can 83 05 wctocﬂ hm L. m> ..v
2535 92° .85 .>_
0252. $252.21:121.1: amazes s.1818262318: <
Ea momma—225 .m> can 832285 .m>
.25 ”an 2:82 2m a~_._~_.o~_§ :2 12 c .25 can. 2:23 3 § 2.: 182232.83
219‘ "and? Gd .2 a? .213 ...wuméd $2 .2 s: m
..qu< ...wumadd 8m. .2 2 a ..an< gum: : : .2 8:
a n a a a a a a a a w .
205.2,: $2 a. a: NE a. 8: “25.26. 8: 2: 8. N: :_ S: o 235 no .85 E
U< van 325.35 .m> O< 9.8 30525 .m> <
.25 as 2:82 2m 3123:2121:c .25 as MESH 2m GS.§.§.§.8_.S:
..huo< d d: $2 .2 mm: 213‘ d £23 82 .2 8: m
..qu< d d<oa :2 .2 8: ..qno< d d: a: .2 :3 2 .2 _ .2 w
noo— -m> onN— .m> :w oro— .m> naN~ .m> new 5: :N p»
22.3 $~E~s amused. 5 2.8: .
. . . . . . . . . . . mm”: 2 0mm 0 you
2325253 .2 22 o: w: e: 3: 2285253 22: we 2: z: .8: < x .ﬁ E E :
.25 as MES»: 2m .2 82.215.223 .25 as 2:83 3 2123221815:
dz...huo< dzﬁuo<
...wuaoﬁué 3am 5. .2 2 : ...dumasaé 0am 82 .2 3: m
dz...$uo< dz...$no< .
...Qumaiozc 02m a: .2 2: ...wnaoiea ”am as .2 8: maﬁa: 2 309mb}.
1245 $5 2 309m:
83 =~ do 86858 $2.8..ch Dan. .3 do 83258 5 2 a ..S c 2 m2 2 m<oo
22:5 “35.25.25 2 as 92.2 : ”magma £205.25 2 $2 222.8: < 25 ”am .6 seem. ._
o< £5.25 2 a: t : o< “85.25 2 $2 8:
”as mass»: 2m .2 a: 12 :2 $2 2.2 a as 2:83 2m .2 $2.832 $2.25
m EOE—=6 mﬁcm—LQQEO m EGEEO mﬂOm—hanac
a U . U a U . U .961— 38:”.—

 

 

vg 3 m: m=o=uow 5:» 03m

 

 

N: 3 2: a:

88m 55 ea

 

 

 

 

Eoﬁtodxm dem 05 no.“ maomtmanu 35A 8% Wm 93m...

~149

Table 56 Example calculation of relative performance (State l-Alligator Cracking)

 

4” AC Thickness

7” AC Thickness

 

 

 

 

 

 

 

 

 

 

Section ID Performance Index Section ID Performance Index
102 90.69 101 23.08
103 25.43 104 3.83
105 95.90 106 16.12
107 21.97 108 70.45
111 65.33 109 75.47
112 10.22 110 52.74
Average 51.59 Average 40.28

 

Mean Performance

(51.59+40.28)/2 = 45.93

 

Relative performance

 

 

51.59/45.93=1.12

 

Relative performance

 

40.28/45.93=0.88

 

 

150

5.5 METHODS FOR INVESTIGATING APPARENT RELATIONSHIP
BETWEEN PAVEMENT RESPONSE AND PERFORMANCE

This analysis is aimed at investigating the relationship between the pavement
responses (deﬂections) and performance measures (cracking, rutting, roughness, etc.).
The usefulness of such relationships can be further divided in two ways:

0 To provide an explanatory information for a given performance trend. For
example a relationship between AC pavement rutting and the farthest FWD
sensor would indicate that rutting is related to the subgrade.

- To provide a predictive capability of the future level of a given performance
measure. For example the initial high average deﬂection of a section may explain

its future cracking and rutting performance.
Relationships were sought by directly relating commonly collected pavement

response data to the development of speciﬁc distresses using statistical analysis. This
analysis was conducted at the site level as well as for the overall experiment. The
following describes statistical techniques used to investigate the apparent relationship

between response and performance measures.

5.5.1 Univariate Analysis

The simpler statistical methods include Univariate and Bivariate analysis of data.
These methods include determination of data statistics such as mean, standard deviation
and data frequencies. Simple histograms and box plots can also be generated to
determine data distribution. These methods also allow for determining the degree of
dependence between variables. The results of such an analysis can be graphically
illustrated. Such an analysis can also provide summary statistics such as the coefﬁcient

of correlation. This analysis was used at the site level. Bivariate analysis can also assist

in identifying outliers.

151

5.5.2 Multiple Regression Analysis

Regression analyses attempt to explain some dependent variable, y, in terms of
many independent (explanatory) variables, x ’s. The model (equation) can be either linear
or non-linear, and with actual, transformed, or interaction clusters of variables. The
model coefﬁcients can be estimated using best (least squares) ﬁtting techniques. The
objective of this method is to develop models explaining the apparent relationships
between pavement performance measures and responses. This analysis was used for

investigating explanatory relationship by using the data from the overall experiment.

152

CHAPTER 6 - ANALYSIS RESULTS FOR THE SPS-1
EXPERIMENT

6.1 INTRODUCTION

The purpose of this chapter is to provide a summary of ﬁndings from the previous
studies and the results of the various analyses conducted for the SPS-1 experiment on
ﬂexible pavements in this study. The performance and structural response indicators used
in the analysis include fatigue (alligator) cracking, rutting, longitudinal cracking in the
wheel path and outside the wheel path, transverse cracking, IRI, and deﬂections
measured by sensors 1 and 7. In this chapter, results from site-level analyses will be
followed by results from overall analysis, and results from apparent relationship between
response and performance, leading to the summary of ﬁndings. Before presenting the
results from analyses, a discussion on the effect of construction on the performance of
SPS-1 pavements and a brief discussion of the performance of test sections at each site

are presented.

The analyses conducted include:

o Site-level analyses (on performance measures): Evaluation of the consistency of the
effects of design factors across sites.

a A comprehensive overall analysis of the performance and response data, which
includes the following methods of analysis:

a) Extent of distress by experimental factors (Frequencies)
b) Linear Discriminant Analysis (LDA)

c) Binary Logistic Regression (BLR)

d) Analysis of Variance (ANOVA)

a An investigation of apparent relationships between response and performance at the
site level and for the overall population.

153

6.2 PREVIOUS STUDIES

This section summarizes the ﬁndings from the literature review of research
reports that deal with pavement performance in the ﬁeld using SPS-l experiment. The
review included FHWA/LTPP reports, NCHRP reports as well as additional literature,
and was focused on research that has identiﬁed factors affecting pavement response and
performance including roughness. The most relevant reports were found to be those from

studies addressing the Long Term Pavement Performance (LTPP) experiments.

6.2.1 Summary of Findings

The information obtained from the literature review has been used to identify
various factors that have been shown in past research as having an effect on response and

performance progression.

6.2.1.1 Factors Affecting Flexible Pavement Performance

A study [1] entitled “Structural Factors for Flexible Pavements” was conducted
using SPS-1 data. The summary of ﬁndings from this preliminary study is given below:

Layer Thickness
o The SPS-1 test sections with thick (178-mm (7-inch)) AC surface layers appear to

be smoother and develop less fatigue cracking than those sections with thin (102-

mm (4-inch)) surface layers. This conﬁrms a similar ﬁnding from earlier studies.

0 In the SPS-1 experiment, AC surface thickness and the age of the project appear
to inﬂuence the amount of fatigue cracking that occurs. The test sections that are

younger and have thicker AC surface layers have the least fatigue cracking.

154

Base Layer

Hot-mix asphalt (HMA) pavements with unbound aggregate base layers show
greater rut depths than those sections with asphalt-treated base layers. This
suggests that a portion of the rutting measured at the surface is a result of
permanent deformations in the unbound aggregate base layer, which is consistent

with a previous ﬁnding from analysis of the GPS test sections.

The HMA pavements with unbound aggregate layers have slightly more fatigue
cracking and higher IRI values than those sections with asphalt-treated base

layers.

The test sections with coarse-grained soils, asphalt-treated base layers, permeable

base layers, thicker bases, and thicker HMA layers were found to be smoother.

The test sections with permeable asphalt-treated base layers exhibit more fatigue

cracking than those without permeable base layers.

Subgrade

HMA pavements built over coarse-grained subgrade soils are smoother than
pavements built over ﬁne-grained subgrade soils. This is consistent with the
ﬁnding in the SPS-2 JPCP: A stiffer foundation contributes to smoother

pavements.

HMA pavements built over coarse—grained subgrade soils and in a no-freeze
climate are smoother and stay smoother over a longer period of time than do those

built over ﬁne-grained subgrade soils in a freeze climate. HMA pavements built

155

over ﬁne-grained sub-grades and in a wet-freeze climate are substantially rougher

than those built in other climates.

HMA pavements built over ﬁne-grained subgrade soils have more fatigue

cracking than those projects built over coarse-grained subgrade soils.

Subgrade soil type and, to a lesser degree, age are important to the amount of
transverse cracking measured at each site. More transverse cracking has occurred
on the HMA pavements built on ﬁne-grained soils than on pavements built on

coarse-grained soils.

Another study [2]was conducted using SPS-1 data to identify the factors affecting

pavement smoothness. A summary of main ﬁndings from this research is given below:

A signiﬁcant difference between early age IRI of pavements placed on DGAB
and ATB was observed. No signiﬁcant difference in early age IRI was obtained
on pavements placed on PATB when compared to other two base types.

The SPS—l projects that showed the highest increase in IRI were located in
Kansas, Iowa and Ohio, reasons being fatigue and transverse cracking for KS
(20), transverse and longitudinal cracking (WP) for IA (19), and rutting for OH
(39) site. Some of the test sections in TX (48) are showing higher increase in IRI
of over 10% within an approximate 6-month period, which is attributed to rutting.
Although the pavements in IA (19), KS (20), and OH (39) achieved a smooth
pavement initially, many sections, including very thick sections had high

increases in roughness during the initial life of the pavement. Achieving a smooth

156

pavement initially does not guarantee that it will remain smooth even during the
initial life.

0 Mix design problems in the AC, inadequate preparation of the subgrade prior to
placing the pavement, or other construction problems can cause smoothly built
pavements to have higher increase in roughness within a short time period.

Two studies[3, 4] were conducted to investigate the effects of sub-drainage on the
performance of asphalt and concrete pavements. The following is a summary of their

ﬁndings for asphalt pavements:

- Based on 7 years (on average) of SPS-l data, those HMA sections built on
permeable bases without edge drains were found to perform better than those with

edge drains.

o The ranking of performance in terms of IRI and cracking for various base types
with all other design features matched is from poor to good performance: un-
drained dense-graded aggregate bases, drained permeable asphalt-treated bases,

and un-drained dense-graded asphalt-treated bases.

0 The results in terms of rutting for the above three sub-drainage designs were
inconclusive.

It was observed from the performance data (see Chapter 4) that a number of
sections located in various sites have shown high distresses (cracking and rutting).
Keeping in view, the early age of the sections, it was felt necessary to investigate effects
of the construction on the performance of all test sections in the SPS-l experiment.

These effects are discussed in the next section.

157

6.3 EFFECT OF CONSTRUCTION ON PAVEMENT PERFORMANCE

For the SPS-1 experiment, detailed construction guidelines were developed by
LTPP (summarized in Chapter 2) for the participating agencies to control variability in
construction across sites. The lesser variability across sites, the lower the “noise”, and
easier it is to determine the “pure” effects of the design factors on pavement performance.
However, some deviations have occurred during construction of various sites and some
of those issues were highlighted in the construction reports prepared by the participating
agencies (see Chapter 3). In general, sections with serious construction deviations will
perform poorer than those with normal construction conditions and inclusion of such
sections in the analysis may distort (bias) the effects of design factors. Sections with
deviations may be identiﬁed at least in three ways, based on:

Construction issues highlighted in construction reports for each site,
0 Unusual performance trend of individual sections, and
0 Unusual material properties of pavement layers.

Depending on the nature of construction issues (construction quality) the
performance of a pavement is affected. Minor issues (such as minor thickness deviation
in base course) may not seriously impact initial performance where as major issues (such
I as HMA mix issues, compaction or drainage problems) have greater chances of affecting
performance early in the life of a pavement. Hence, a construction deviation (poor
quality) may be used to identify sections that may potentially show an abnormal
performance. Also, some sections with serious early performance concerns were “de-
assigned” from LTPP database (for example, 39-101 and 20-101).

In this study, any abnormality in early performance was used as an indicator to

identify substandard sections. The performance of all the sections, over time, was

158

observed for this purpose and those sections that had premature “failure” (within ﬁrst 2 to
3 years of service life) were identiﬁed. It was observed that most of the identiﬁed
sections are from a few sites (for example KS for fatigue cracking and TX for rutting in
SPS-l) in the experiments, indicating consistent construction problems in those sites. In
such cases, all sections from the identiﬁed sites were excluded from related analyses.
Material properties (example: HMA mix properties) of pavement layers may also indicate
non-compliance in construction. Limited material data are available in DataPave (Release
17). This data were also considered to explain the probable causes of unexpected
performance wherever possible.

In order to further investigate the construction-related performance issues, each
performance measure for all pavement sections in SPS-l experiment was examined over
time. This analysis helped minimize the bias, if any, in the results. The analysis is
discussed next with illustrations. This section of the report is followed by site-wise

performance summaries of the test sections.

6.3.1 Construction-related Issues

A brief discussion of construction-related performance issues for each
performance measure in SPS-l is presented in this section of the report. Based on the
time—series plots for all distress measures it was found that premature “failure” was
predominantly observed in rutting for a relatively large number of pavement sections.
Hence, this is presented ﬁrst among all the performance measures.

Rutting
Figure 6- 1- shows rutting in all the SPS-l test sections over time. It can be

observed that a considerable number of sections have noticeably high initial rutting. The

159

premature rutting in these pavements can be further classiﬁed into two types based on the
causes of rutting:
- Mix-related rutting, and
a Base layer rutting (this could be because of wet base and poor drainage or
poor compaction in un-bound pavement layers).

Therefore, the premature or early rutting in pavements was separated from the
structural rutting in order to study the effects of structural factors on the long-term
pavement performance. Table 5-1 is a summary of details regarding the pavement
sections that exhibited early (pre-mature) rutting in the SPS-1 experiment. The causes of
rutting were identiﬁed by using-the transverse proﬁle data available in the LTPP database
based on the criteria developed in NCHRP Report 468 [5]. Figure 6-3 through Figure 6-6
show the average transverse proﬁles of some sections, from four of the SPS-1 sites,
which showed premature rutting. HMA material-related data ﬁom the ﬁeld cores were
extracted from DataPave (Release 17.0). Unfortunately a limited amount of material data
is available at this point in time; therefore, only a few of the mix-related properties could
be calculated. The summary of the mix-related data for the sites that showed early rutting
is given in Table 6-2. It can be seen from these asphalt mix properties that there is high
variation in the ﬁeld air void content between these identiﬁed sites. High air voids in the
pavement sections at the Kansas site KS (20) and very low air voids (high VFA) at the
Texas site TX (48), are noticeable. The pavements built at these two sites have shown
extensive cracking and rutting, respectively.

To investigate the effect of structural factors on the rutting performance, the
pavement sections were separated, as explained above, into two categories: (a) pavements

with premature rutting, and (b) pavements which exhibited structural rutting. Figure 6-2

160

shows the rutting performance of sections with probable “structural” rutting. The effect
of outliers (sections with premature rutting) on the rutting performance for SPS-1
pavements can be observed by comparing Figure 6-7 and Figure 6-8.

An analysis of variance (AN OVA) was performed separately on two subsets of
the data as well as on data from all the sections (superset of the two subsets). The ﬁrst
subset represents the pavement sections which have shown higher rutting at an early age
(mix-related or material-related). The second subset includes the pavements which have
shown normal rutting growth, and these pavements were assumed to be exhibiting
“structural rutting”. Table 6-3 is the summary of results from AN OVA. Initially, only the
main effects of structural factors were considered in the analysis by blocking the site.
The results indicated that none of the structural factors has a signiﬁcant effect on
premature rutting. These results are reasonable, as it is expected that pavements will
undergo accelerated rutting early in their service life (irrespective of the pavement
structure) if the asphalt layer has mix-related issues or when the base has drainage-related
issues. Next, ANOVA was performed by taking all the experimental factors and the
results are shown in Table 6-5. The mean rut depths from both analyses by each
experimental factor are shown in Table 6-4 and Table 6-6, respectively, to illustrate the
effects. A brief discussion of the results from analysis of structural rutting is given
below:

HMA Thickness: Pavement sections with “thin” (4-inch) HMA surface layer have
undergone higher rutting compared to those with “thick” (7-inch) HMA surface layer.
However, this difference was not found to be of practical signiﬁcance at this point in

time.

161

Base Type: No statistically signiﬁcant effect of base type was observed on the structural
rutting, at this point. On average, sections with DGAB have shown slightly higher rutting
than those built with treated bases. A slight (0.05<p-value <O.1) interaction effect was
observed between base type and subgrade type. Among pavements built with DGAB,
those built on ﬁne-grained soils have shown higher rutting than those built on coarse-
grained soils.
Base Thickness: Higher rutting occurred in sections with 8-inch base than those with 12-
inch or l6-inch base. However, the effect is statistically marginally signiﬁcant, is not of
practical signiﬁcance. .
Drainage: On average, sections with no drainage have shown slightly higher rutting than
those with drainage. However, this effect of drainage on the structural rutting was not
found to be statistically signiﬁcant at this point in time.
Subgrade Type: Rutting in sections built on ﬁne-grained soils is comparable with rutting
in sections built on coarse-grained soils. The effect of subgrade types was not found to
be statistically signiﬁcant for structural rutting.
Climatic Zone: Effect of climate was found to be statistically signiﬁcant. Pavements
located in WNF zone have shown higher rutting than those located in WF zone.
However, this effect is not of practical signiﬁcance.

Given the above ﬁndings, all subsequent statistical analyses on rutting
performance, (presented later in this chapter) were performed on data from the sections
with structural rutting. This assumes that only pavement sections which have exhibited

structural rutting will capture the effects of design factors on the long-term rutting

performance.

162

Table 6-1 Identiﬁed sites and sections with rutting problems

 

Site

Sections deleted due to
extensive rutting

Probable cause
of rutting

Comments

 

Arizona, AZ (4)

All 12 sections

HMA and Base

The rutting is either
occurring due to HMA
mix problems or base
or both (from

transverse proﬁle)

 

Kansas, KS (20)

All 12 sections

HMA and Base

Some sections have
shown HMA mix
related rutting and
others showed base
problems (from

 

Michigan, MI (26)

113, 114, 119, 122

120

HMA

transverse proﬁle 2

The ﬁrst four sections
were deleted from the
LTPP database just
aﬁer construction

 

Nebraska, NE (31)

All 12 sections

HMA

From transverse
proﬁle

 

Ohio, OH (39)

101,102

Base

From transverse
proﬁle

 

Texas, TX (48)

All 12 sections

HMA

See reference[6]

 

Virginia, VA (51)

113

 

Base

 

Only one section has
shown accelerated
deterioration at this
site due to base
problems (from

transverse proﬁle)

 

Total

 

 

56 sections

 

 

Table 6-2 Average asphalt mixture properties in the ﬁeld

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

Asphalt Air Void a a Superpave Speciﬁcations
State Cogent Cogent V23: ‘23:; VM A1 VFAZ
( °) ( °) on) (%)

4 4.4 10.6 19.1 44.3

19 4.2 10.5 - -

20 4.1 15.3 21.4 29.1

26 5.0 6.5 - -

31 4.2 6.2 - - >14 “'75

39 6.6 11.2 - -

48 4.4 1.8 12.2 85.0

51 4.9 9.7 20.4 52.6

 

 

 

 

a . . . 1 . .
Note: Gsb values are m1ssrng therefore, these propertres can not be calculated, the minimum VMA

2
requirement for nominal maximum size of 12.5 mm, VFA requirements for traffic > 100 million
ESALs (Source: Superpave Mix Design, SP-Z)

163

Rut Depth (mm)

35

W
O

N
kl!

N
O

15

Rut Depth (mm)

 

   

     
 
 

WW 7"“ pic“ __—. '
”As-ff- ”15:3- ‘?-v'

'1’ p " a ‘7 «6"' *—
:%X%£ . ”,3:

—=?- _ gat~ .
--.;.,-m r 2—..

’5 a: ._ _ ~ .

 

 

o 1 2 3 4 5 6 7 8 9 10
Age (years)
Figure 6-1 Rutting with time for SPS-l pavements - All sections
l
t
t
l
l
i
.1

 

 

Age (years)

Figure 6-2 Rutting with time for SPS—l pavements — Selected sections

164

 

 

2 51— m cocoom lmEzE A038 Sm 01.35 6205:»; v-0 2sz

9:5 omen 2: Eat 353.5

 

 

 

2 3-; cozoom lchE :madma c8 058d 089235. v-9 oBmE

955 own; 2: 80¢ 353me

II I ll .l1liilllllll l1 ll;

at- . . a

 

 

 

 

'l,l1 o”-

(urtu) qidap mg

(W) tlldap m‘d

851mm eozoom [um—HEP. A093 + 5):.5 no.“ oEEd new—05:29. m6 05m?—

AEEV once 05 Eat cue—22D

dlilii.|1 l l l|1| ' ill

(urur) qidop Ina

 

 

 

85-8 cozoomlwiza umun com 0595 3555;. m6 823.;

AEEV awe... of an 8:820

Eilllliiilllu

(unn) qidap in};

 

 

 

Rut Depth (mm)

Rut Depth (mm)

35

30

25

20

15

10

35

30

25

20

15

10

 

 

y = 3.623Jc0'27

 

 
    
 
    
 
 

 

L- “ 2—03
R— . 5
o o
o o o
o o o
- o o
no. 0 o o o
o o
o o o
- o coo. o o o
o o
oo o o o
o o o o oo o oo o
no 0 o o oo o I. an o o a
1— o o o c o o o a on o 00. no.0
no. oo o o a.-. can 0 no... on. o 00 l
i o o o o oaoco a oo-oooo on a o a one. 00.. c I
oo o o o no I 00.00.. no -99 co. co .0.. on one- o o
oo o o. o o 0000 coon. - alt-coo o o
l- on. on... m.- cumo o o o 0
mo.- -o-ooooomnuo—umo. .- 000 co
co co..." -0... o a.-. can. can: 00 o o o o
(— coo- uoo .- a 00 0 on
I L 1 1 l 1 1 1 L 1 1 . L - l 1 L 1 1 L L

 

2

4 6 ' 8 l 0 11
Age (years)

Figure 6-7 Rutting growth with time for SPS-l pavements - All sections

 

.._-.I————

 

P ' 0.285
y=3.27x

. R2=0.39

).

l.

L . ' ° : ..

l. z: .

‘ ::. .°..... °. °.:::°...:. °...°. :..... ::..' °

 

 

 

  
 
    

 

 

4 6 8 10 12
Age (years)

Figure 6-8 Rutting growth with time for SPS-l pavements — Selected sections

166

Rut Depth (mm)

Rut Depth (mm)

35

3O

25

20

15

10

35

3O

25

20

15

10

 

1 ° R2=0.35

O O
O O O
O O O
- O O
O
O OO O O O O
O O O
O O O
t- O OO O I O O O
O O
O O O O O
O O O O O O O O O O
CO O O O O OO O O O ” O O Q
- O O O O O O O O O O O O O O O OO O O
O O. OO O O O O- O O O D O . O O O O OO O O OO
O O O O OOOC. O O O- “OO o. a O - ”O- O 0.0 O
.
.O C O O OO . cameo- an -OD OOO OO -.O “ONO- O
OO O O. O O OOOO .O0-0 - ”-NO” O O
'- OCO OO .0. o- .0“. -. O- - O O . O O
“OOO- -.-OCOUO .- .- -C —. “O. O . O OO OO
4 ﬂ - C 0.0 -..OO O ~- . OO ”O coca OO O C O O

 

l 1 L 1 l L 1 4 1 I 1 1 ' l 1 L 1 1 l

 
    
 
   
 

0.27
X

 

0 2 4 6 8 10
Age (years)

Figure 6-7 Rutting growth with time for SPS-1 pavements — All sections

 

L y = 3.27x
R2=O.39

 

 

 

   
 
    
  

O
l.
O
l- O O O
O O O O
O O OO O O O O
*- O O O O O O O O O O O O O O O
O O O O . OOO O O O O C O O OO O O
O O O ca 0. O C O- ”OO O on O - O O. OCO
OO O O O Q. O. 0. -OO OOO OO -OO “ONOO O O.- O
OO OO O O OO O...“. - co- OO” “0.0.0 In - - .
OO OO.“ - "O. . .3. O-- O O .
-0.-CO a. on. “- -. .0“. O O O OO O
. . OCO -..OO O a. C - OO “O OO“ O O O O O
.O OO COO .O C O O O O a
1 L 1 1 1 1‘ 1 1 1 1 l 1 1 1 L

f 0.285

 

0 2 4 6 8 10
Age (years)

Figure 6-8 Rutting growth with time for SPS-1 pavements — Selected sections

166

 

 

 

Table 6-3 Summary of p—values from AN OVA for determining the effect of main design
factors on pavement rutting

 

 

 

 

 

 

 

 

 

 

 

 

Rutting Type
Desrgn Factor NOD-S®C?fal Structural Overall
ruttmg ruttrng
HMA thickness 0.71 0.074 0.20
Base type 0.20 0.51 0.017
Base thickness 0.99 0.08 0.195
Drainage 0.12 0.25 i 0.030
Site (blocked) 0.15 0.00 0.00
R2=0.343 1820.55 RT=0.57
N=53 N=159 N=212

 

a . .
Note: Mix-related or premature rutting 1n un-bound layers.

Table 6-4 Summary of marginal means from ANOVA for determining the effect of main
design factors on pavement rutting

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

Rutting Type
Desrgn Factor Non-structural Structural rutting Overall (mm)
(mm) (mm) -

HM A 4” 9.0 5.3 6.1
mime-95 7” 10.0 4.9 5.8
DGAB 1 1.0 5 .2 6.5
Base type ATB 9.05 4.9 5.7
ATB/DGAB 8.2 5.1 5.6
8” 10.0 5.6 6.3

Base ,,
thickness 12 10-0 5-1 5.8
16” 9.05 5.0 5.8
N 11.0 5.3 6.3

Drainage

Y 9.0 4.9 5.6
MSEa 0.206 0.062 0.10

 

a
MSE is in natural log.

167

 

 

Table 6-5 Summary of p-values from AN OVA for determining the effect of experimental
factors on pavement rutting

 

 

 

 

 

 

 

 

 

 

 

E , 1 F Rutting Type
xpenmenta actor Non-structural ruttinga Structural rutting
HMA thickness 0.16 0.043
Base type 0.94 0.54
Base thickness 0.76 0.09
Drainage 0.50 0.28
Subgrade 0.46 0.43
Zone 0.27 0.00
Trafﬁc 0.000 0.013
R2=0.552 R2=0.55
N=53 N=159

 

 

 

 

a . . .
Note: Mix-related or premature ruttmg 1n un-bound layers.

Table 6-6 Summary of marginal means from ANOVA for determining the effect of
experimental factors on pavement rutting

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

Rutting Type
Design Factor
Non-structural rutting Structural rutting
HMA 4” 9.7 5 .7
thickness 7” ' 11.8 5.0
DGAB 10.7 5.4
Base type ATB 10.7 5 .2
ATB/DGAB 10.7 5.3
Base 8” 9.7 5.7
thickness 12” 10.7 5 .0
16” 10.7 5.2
Drainage N l 1.8 5.5
Y 10.7 5.1
F 13 5.2
Subgrade C 9.7 54
WP 6.5 4.9
Zone WNF 13 5.6
DF 14.4 4.0
DNF - 7.1
MSEa 0.137 0.091

 

 

 

a
MSE is in natural log.

168

 

Fatigue Cracking

It was observed that sections from the Kansas, KS (20), site exhibited the highest
area of cracking at an early age compared to sections from other sites. The KS (20) site
had a wet subbase during construction (based on the construction report). Also, from the
materials data (DataPave) it was found that the test sections at KS (20), on average, have
“high” air void content in the HMA (see Table 6-2). These reasons could have caused the
abnormally high cracking in the sections at this site. Figure 6-9 and Figure 6-10 are time-
series plots of fatigue cracking for all the pavements sections, before and after exclusion
of sections from the Kansas site, KS (20). All statistical analyses pertaining to fatigue
cracking, presented in this chapter, were conducted without including data fr0m sections

at Kansas site, KS (20).

Roughness and other Performance Measures

Figure 6-11, through Figure 6-14 show the time-series plots for IRI, transverse
cracking and longitudinal (WP and NWP) cracking, respectively. It can be observed that
only a few sections have exhibited an abnormal performance. Exclusion of data ﬁom
these sections was not considered necessary as their inclusion will not impact the results
considerably. Therefore, all the pavement sections were included in the analyses of

roughness, longitudinal cracking (WP and NWP) and transverse cracking.

169

3%

 

b.)
U1
0

8
o

F‘ N
O U1
C O
'V—rrr—r—r—T-r—r—r-r r-r-r'T" | -r—1—r"r-1 ﬁ-Tw—r—r-Tﬁ-r'r—T r—r—r—r r-T-vﬁ—r

 

Fatigue Cracking Area (sq-m)
N
o
o

U!
C

 

 

O

‘4

O
r—b
N
b)
.h
U‘
CA
\I
m
\0
~11
O
H

Age (years)

Figure 6-9 Fatigue cracking with time for SPS-1 pavements — All sections

 

400

350

Fatigue Cracking (sq—m)
- H N N b3
8 8 8 8 8
O H'r—r—r'r—r‘r'r‘r-r—r 1—1‘1—1 T—r‘ﬁ‘rTT‘r—T—r—r‘T—r—r—rT-w-‘rn *rﬁ—r'r-r‘r—j

U1
0

\
:1

)1
\\\

 

 

 

.1/13' )2"
ii)“ I] 1"
l/ :44» «a.
o 2.}. z» M~<~- ”2...... ._
2 4 6 8 10 1

Age (years)

Figure 6-10 Fatigue cracking with time for SPS-l pavements — Selected sections

170

s» .e
'03 U1 A U1 U1
“H‘- I+I-+Y—*FT“4_TT ' "l

IRI (tn/km)
:9
U"

2

 

80

70

Transverse Cracking (m)
N L» a U1 o
o o o o o

H
O

 

 

 

 

 

 

 

 

 

0 l 2 3 4 5 6 7 8 9 10 l 1

Age (years)
Figure 6-11 IRI with time for SPS-1 pavements - All sections

_ 1

.

L

l

l

11— — -1«-—. — 3

0 l 2 3 4 5 6 7 8 9 10 1
Age (years)

Figure 6-12 Transverse cracking with time for SPS-l pavements — All sections

171

 

Long. Cracking-WP (m)

Age (years)

 

Figure 6-13 Longitudinal cracking—WP with time for SPS-1 pavements — All sections

 

 
    
   

    

 

350 f
300 4;
E 2501» / /.'/" ‘
5 _. 1 11
Z 200T / 7 ll], 9
'2’ 150l -—/572,,I: ~ kgéx4géV—z—é‘; 1‘
“030 i ll" / 7 '2."
.8000? V//,' ,1... i \\ r“
T ,1 x- " . .22., \1 /« fs /
3. I’M "' I, ‘\ .. 1’ \\V "
50 , 01y; / - 9M \9 ' .
l. [1% If.” y): 5:131.” \‘; _,/,.»/” 1
01 , .-' :.- :1_. 1
0 1 2 3 4 5 6 7 8 9 10 1
Age (years)
Figure 6-14 Longitudinal cracking-NW? with time for SPS-l pavements — All sections

172

6.3.2 Drainage-related issues

In the above section, construction-related issues have been linked to the poor
performance of some pavement sections. Some construction and/or maintenance related
issues with respect to the in-pavement drainage were also identiﬁed in previous research
[3, 4]. The in-pavement drainage for some of the SPS-1 ﬂexible pavement sections was
found to have some deviations from design. All the drained sections of the SPS-1
experiment were video taped to assess the condition of the drainage in the project 1-34C
[4]. A subjective assessment of the quality of the drainage functioning in each test
section as “good” or “poor” was reported. The ratings assigned to each section are
summarized in Table 5-7. The “poor” rating was indicative of; (i) buried lateral outlet,
(ii) outlet fully blocked with silt, gravel or other debris (iii) longitudinal drains being
fully blocked, or (iv) a considerable amount of stagnant water in the longitudinal drain.
A “good” rating was given to the drainage if a reasonably sufﬁcient ﬂow of water was
. evident even if some amount of material was present in the drains. Hall et al [4]
conducted preliminary analysis of the performance of SPS-1 test sections in light of their
assessment of drainage, and a brief summary of their ﬁndings are presented below:

- Undrained pavement sections built on DGAB may develop cracking, rutting

and roughness more rapidly than drained sections built on ATB.

o Undrained pavement sections built on ATB may develop roughness and

cracking more slowly than those built with drained DGAB, while the un-
drained sections may develop rutting more rapidly.

o Undrained pavement sections built on ATB/DGAB may develop roughness
and rutting more quickly than those on drained DGAB, while the undrained
sections may develop cracking more slowly.

0 Also, among the drained sections, those with “good” rating for drainage
performed better than undrained sections, while those with “poor” rating did
not.

173

However, the above trends were based only on the average performance and in no
case, were the differences detected statistically signiﬁcant. These preliminary ﬁndings
(from Hall et a1) should be considered during the interpretation/validation of the results ‘

(from this study) regarding the effect of drainage.

The results from the site level analysis for SPS-1 experiment are summarized in

the next section.

Table 6-7 Subjective ratings of drainage functioning at SPS—l test sections based on
video inspection results (source: [4])

T Section

KS (20)

 

G= Drainage function rated as good
? = Drainage outlet not found

P = Drainage function rated as poor
?'= Camera could not be inserted

174

6.4 SPS-1 PROJECT PERFORMANCE SUMNIARIES

This section summarizes the performance trends for each site within the SPS-l
experiment based on the latest year data. The performance summary for each site is based
on the data available in the Release 17.0 of the DataPave. The severity levels for all
types of cracking were combined to calculate its total magnitude. This descriptive
summary is intended to help the reader gain an understanding of performance of test
sections at each site. The performance of pavement sections regarding selected distresses
is presented here, for each site. The identiﬁed distresses include fatigue cracking (sq-m),
longitudinal cracking-WP (m), longitudinal cracking-NW? (m), transverse cracking (m),
rutting (mm) and roughness (m/km).

Additional details about each of the sites can be found in site-level summaries

presented in Appendix A1 and performance data tables in Appendix A2 of reference [7].

Alabama, AL (1)

Performance data is available for 10 years (1994-2003) at this site. Fatigue
cracking is the dominant distress type at this site. Section 103, 104, 106, 107 and 112
have less than 10% (area) cracking while all other sections have fatigue cracking of range
10% to 15%. A wide range of longitudinal cracking-WP (between 5 m and 30 m)
occurred on all the sections. Longitudinal cracking-NWP, between 80 m and 200 m,
occurred on all the sections, by year 8. Transverse cracking, of range 15 m to 50 m, was
observed in sections 101, 102, and 105 respectively. Sections 102 and 105 have shown 10

mm and 17 mm of rutting, respectively, while other sections have rutting between 6 mm

175

to 9 mm. Sections 102 and 107 have IRI of 1.4 m/km and 1.7 m/km, respectively, while
other sections have IRI less than 1.0 m/km.
Arizona, AZ (4)

The performance data is available for 10 years (1994-2003) at this site. Less than
3% (of area) of fatigue cracking occurred in sections 113, 119 and 124, while less than
1% cracking occurred in other sections. Longitudinal cracking-WP is the dominant type
of cracking with all the sections showing a cracking between 10 to 150 m. Longitudinal
cracking-NWP of 150 m and 120 m occurred on sections 114 and 120, respectively;
while in other sections longitudinal cracking-NWP is less than 50 m. Transverse
cracking of 76 m and 45 m occurred on sections 113 and 121, while other sections have
less than 30 m of transverse cracking. Rutting of 14 mm and 25 mm occurred on sections
114 and 119, respectively, after 6 years. In other sections rutting ranged from 3 mm to 9
mm. All sections except 113, 120 and 122 have IRI greater than 1 m/km while other
sections have IRI less than 1.0 m/km.

Arkansas, AR (5)

The performance data is available for 9 years (1995-2003) for this site. Fatigue
cracking area ranged from 10% to 25% in sections 119, 120 and 121; whereas, all other
sections have exhibited less than 10% of fatigue cracking area. All the sections exhibited
longitudinal cracking-WP less than 10 m. All the sections have exhibited longitudinal
cracking-NWP, which range from 140m to 280 m. Transverse cracking of 48 m was
observed only in section 119, whereas all other sections have less than 20 m of cracking.
Rutting between 5 mm to 9 mm was observed at the site. Sections 119 and 120 have IRI

of about 1.7 m/km while other sections have IRI of about 1 m/km.

176

Delaware, DE (10)

The performance data is available for 7 years (1996-2003) for this site. Fatigue
cracking is the dominant distress type at this site. Sections 101 and 102 have cracking of
about 10% and 20%, while in other sections cracking was less than 10%. Longitudinal
and transverse cracking did not occur on any of the sections. All sections except 102 have
shown rutting of range 2 to 4 mm while, sections 102 has 7 mm of rut depth. IRI for all
the test section is less than 1.0 m/km.

Florida, FL (12)

The performance data is available for 7 years (1996-2003) for this site. Fatigue
cracking less than 1% occurred in the sections. Longitudinal cracking-WP, less than 10
n1, occurred in sections 107, 108, 110, and 112. Longitudinal cracking-NWP, less than 50
m was observed in sections 101,105,108 and 110. Transverse cracking was not observed
on any of the sections. Rutting of about 4 mm occurred in all sections, except sections
103, 105, 110 and 111, which exhibited rutting of about 6 mm. All the test sections have
shown less than 1 m/km of IRI.

Iowa, IA (19)

The performance data is available for 9 years (1995-2003) at this site. Fatigue
cracking of 10% (area) was observed on section 102, and all other sections have fatigue
cracking less than 2%. Longitudinal cracking-WP of range 50 m to 100m occurred in
102, 104, 105, and 107, while in other sections this cracking is less than 30 m.
Longitudinal cracking-NWP of range 110 m to 295 m occurred at this site, in all the
sections. Transverse cracking of 30 m to 70 m was also observed in sections 101 through

106, while in other sections this cracking is less than 20 m. Sections 107, 108, and 109

177

have rut depth of about 7 mm while other sections have rutting between 3 mm and 6 mm.
Sections 101 and 102 have IRI values of 1.8 ndm and 2.5 m/km, while IRI in other
sections ranged from 1.0 m/km to 1.6 m/km.

Kansas, KS (20)

The performance data is available for 8 years (1993-2001) at this site. Fatigue
cracking is the main distress type in all the sections. Fatigue cracking ranged from 5% to
20%. Longitudinal cracking-WP has not occurred at this site. Sections 103, 104, 105 and
110 have longitudinal cracking-NWP of 72 m to 189 111, while in other sections the this
cracking is less than 20 m. Sections 103, 105 and 110 have transverse cracking less than
16 m. Sections 101, 102, 107 (data available for ﬁrst 2 years only) have shown rut depth
between 16 mm to 25 mm, whereas section 105 has exhibited 13 mm of rutting, after 3
years. Other sections have rutting less than 5 mm. Section 105 has an IRI of 2.7 m/km
while sections 104, 109 through 112 have IRI less than 1.2 m/km. Other remaining
sections have IRI between 1.5 and 2.0 m/km.

Louisiana, LA (22)

The cracking data is available for only 2 years (1997-1999) at this site. The
rutting data is available for 6 years (1998-2003) while roughness data is available only
for one year (1997). No cracking occurred on any of the sections. Sections 119, 122 and
123 have rutting less than 4.0 mm while; other sections have rutting of about 5 mm. All

sections have initial roughness less than 0.8 m/krn.

178

Michigan, MI (26)

The performance data is available for 7 years (1996-2003) for this site. The
performance data is only available for eight sections for this site, as four sections (112,
114, 119 and 122) have been ‘de-assigned’ from LTPP due to construction issues.
Fatigue cracking of 2 to 10% was observed on sections 115, 116, 117 and 124, while no
cracking occurred in other sections. Longitudinal cracking-WP and transverse cracking
has not occurred on any of the test sections. Longitudinal cracking-NWP was observed
on all test sections. Section 116 has longitudinal cracking-NWTJ of 35 m while this
cracking in other sections ranged from 120 m to 188 m. Sections 115 and 117 have
rutting of 9 mm and 12 mm while other sections have rutting of about 6 mm. Sections
118 and 117 have IRI of 1.2 m/km and 1.4 m/km whereas other sections have IRI less
than 1.0 m/km.

Montana, MT (30)

The performance data is available for 5 years (1998-2003) at this site. Fatigue
cracking was observed in sections 115, 117, 120, 121, and 122, with a range of 5% to
10%. Other sections have fatigue cracking of less than 5%. Longitudinal cracking-WP
occurred only in sections 113, 114, 115, 118, and 124, with a range of 5 to 10 m.
Longitudinal cracking-NWP occurred in all sections, with a range of 150 to 208 m.
Transverse cracking were observed Only in sections 113, 115 and 121, and this cracking
was between 5 m to 10 m. Rutting of 8 mm occurred in sections 120 and 121, while it
ranged from 3 to 5 mm in other sections. Sections 120 and 121 have IRI of about 1.5

m/km whereas other sections have IRI of range 0.8 m/km to 1.1 m/krn.

179

Nebraska, NE (31)

'The performance data is available for 7 years (1995-2002) at this site. Fatigue
cracking has just initiated only in sections 113 and 114. Longitudinal cracking-WP of 97
m was observed in section 113. Longitudinal cracking-NWP and transverse cracking did
not occur in any of the sections. Rutting of 29 mm was observed on section 113 by year
5. Among other sections, 114, 115, 118, 123, and 124 have rutting between 11 mm and
15 mm, while others have rutting between 5 mm and 8 mm. Sections 113 has an IRI of
1.9 m/km and all other sections have IRI of about 1.0 m/km. All test sections have an
initial IRI between 0.9 and 1.4 m/km.

Nevada, NV (32)

The performance data was collected for 8 years (1996-2003) for this site. Fatigue
cracking area is less than 1% in all the test sections. No noticeable longitudinal cracking-
WP was observed in any of the sections. More than 50 m of longitudinal cracking-NWP
have been observed only in section 103 and sections 101, 102, 104, 105 and 109 have
less than 35 m of longitudinal cracking-NWP. Less than 15 m of transverse cracking
occurred in sections 102, 107 and 109. Sections 101, 104, 107, and 108 have rut depth
less than 4 mm while other sections have rut depth of about 5 mm. All sections except
102 have IRI less than 1 m/km. Section 102 has an IRI value of 1.4 m/km.

New Mexico, NM (35)

The performance data was collected for 8 years (1997-2003) at this site. Fatigue
cracking, less than 1% of area, occurred in sections 102 to 104. Sections 101, 103,105
and 107 have exhibited less than 45 m of longitudinal cracking-WP, while section 102

has 70 m of this cracking. Furthermore, less than 25 m of longitudinal cracking-NWP

180

was observed only in sections 101, 102, and 107. No transverse cracking was observed at
this site. All sections at this site have exhibited rut depth of about 7 mm. All sections
except 101, 102, 103, 107 and 112 have IRI between 1.0 to 1.3 m/krn while other sections
have less than 1.0 m/km of IRI.

Ohio, OH (39)

The performance data was collected for 7 years (1996-2002) for this site. At this
site, cracking data for sections 101, 102,105 and 107 are only available for the initial
year. These sections were ‘de-assigned’ from the LTPP because of premature rutting.
Fatigue cracking is the dominant cracking distress within all sections at this site. All
sections, except 109, have fatigue cracking between 3 to 15% of area. All sections,
except 104, 111, and 112, have between 175 to 245 m of longitudinal cracking-WP. Also,
all sections except 109, 110, 111 and 112, have longitudinal cracking-NWP between 200
to 260 m. No transverse cracking was observed in any section at this site. Sections 101
and 102 had more than 10 mm of rut depth after only 1 year. Also sections 103 108, and
109 have exhibited rutting of about 10 mm. IRI more than 2 m/km was observed on
sections 101 and 102 after only 1 year of construction. All sections have IRI greater than
1.4 m/km, while sections 103 and 108 have IRI of 3 and 2 m/km, respectively.
Oklahoma, OK (40)

The performance data was collected for 6 years (1997-2003) for this site. Less
than 3% (area) of fatigue cracking was observed in all the sections. No longitudinal
cracking-WP occurred on any section at this site. Longitudinal cracking-NWP between
20 m and 120 m was observed in all sections except, section 113, which has no

longitudinal cracking-NWP. Transverse cracking less than 4 m was observed in sections

181

113, 115 and 121. Sections 113, 117 and 122 have more than 10 mm ofrut depth, while
the other sections have rutting between 4 to 8 mm. Sections 113 through 118 have IRI of
about 1.1 m/km whereas, all other sections have less than 1.0 m/km of IRI.

Texas, TX (48)

At this site, performance data was collected for 5 years (1997-2002). Fatigue
cracking, less than 1% (area) and longitudinal cracking-WP (between 20 m and 50 m),
was observed only in sections 113, 117, 118 and 122. Only sections 112 and 119 have
exhibited longitudinal cracking-NW? of 54 m and 77 m, respectively. No transverse
cracking was observedin any of the sections. Severe early rutting, between 10 mm to 18
mm, was observed in sections 114, 115, 116, 119, 123 and 124, after 1 year. Rutting in
these sections progressed to about 14 mm to 26 mm, by year 4. All other sections have
rutting less than 11 mm, by year 4. Sections 115, 116 and 119 have IRI between 1.2 and
1.8 m/km, after only 2 years and all other sections have IRI less than 1.0 m/km.
Virginia, VA (51)

The performance data was collected for 6 years (1995-2002) for this site. More
than 25% of fatigue cracking was observed only on sections 103 and 120. Longitudinal
cracking-WP and transverse cracking has not occurred on any of the sections. Sections
114 and 120 have exhibited longitudinal cracking-NWP of 40 m and 4 m, respectively.
Section 103 has rut depth of 12 mm just after 1 year and this progressed to 21 mm by
year 5. All other sections have exhibited rutting ranging between 4 mm and 6 mm. All

sections, except 113, have IRI less than 1.3 m/km. Section 113 has an IRI of 1.9 m/km.

182

Wisconsin, WI (55)

The performance data was collected for only 4 years (1998-2002) at this site.
Fatigue cracking less than 5% was observed only on sections 113, 114, and 116.
Longitudinal cracking-WP and transverse cracking has not occurred on any of the
sections. A wide range (between 90 m and 300 m) of longitudinal cracking-NW? was
observed on all the sections. Sections 120 and 121 have rutting of 4.0 mm while all other

sections have rutting between 6 to 9 mm. All sections have IRI less than 1.3 m/km.

6.5 SITE-LEVEL ANALYSIS

The site-level analysis deals with each SPS-l project separately. The main
advantage of this analysis is that it eliminates the variability between SPS-l sites. For
each site, the climatic conditions, subgrade type and trafﬁc are the same. Construction
conditions, material sources and surveys can also be considered the same for a given
SPS-1 project. Two approaches were used: (1) paired comparisons by design factor; and
(2) evaluation of the various designs within the project.

As described in Chapter 5, the site-level analyses consists of two types of
comparisons: (i) Level-A— In this analysis all designs (0101 through 0112, or 0113
through 0124) at a given site are compared such that only one factor is held common
within the sections of each group. For example in level-A analysis, the effects of HMA
thickness (4” vs. 7”) were considered for all twelve sections within a site by ignoring
base type & thickness, and drainage. (ii) Level-B—analysis- In this analyses, most of the
factors are ‘controlled’ for comparisons. In level B analysis, the effect of HMA thickness

(4” vs. 7”) was compared for only those sections for which all other factors are the same.

183

The concepts of Performance Index (PI) and the Relative Performance Index (RPI) were
used in the site level analysis. PI and RPI were introduced in Chapter 5 of this report. The
site-level analysis process is summarized in Figure 6-15. Each analysis was conducted
separately for each performance measure. The pavement performance measures
considered include:

Fatigue cracking

Rutting

Roughness (IRI)

Transverse cracking, and
Longitudinal cracking (WP and NWP)

The P15 and relative performance ratio for the main design factors were calculated
at all sites. Because the relative performance is a ratio, it can be used across the sites. The
relative performance ratios for a given design factor from all eighteen states were used to
test the signiﬁcance of its effect. A two-level non-parametric (W ilcoxon Signed Ranks)
test was done to evaluate the effect of HMA thickness and drainage, and a multiple level
non-parametric comparison test was done to evaluate the effect of base type. The p-
values from these tests are reported in the discussion of the results. To evaluate the
interactive effect of the design factors with climatic zone and subgrade type, the average
relative performance ratios within the same climatic zones and for both subgrade types
were compared.

The computed P15 and relative performance ratios for all the distresses are
summarized and presented in Appendix A4 of reference [7]. The following is a summary
of the main ﬁndings from each method of analysis, categorized by design factor and

performance measure.

184

 

Site Level Analysis

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

1
[ Level-A Comparisons ] [ Level-B Comparisons ]
................................... ,'""""""""""""""""""‘1
: E i ~ 5
E Effect of : 5 Effect of4Hl:’/ISA7T:h1ckness :
i HMA,Th1ck,ness : i Controlling for other factors E
: 4 VS. 7 : : |
l E } Effect of Base Type 1
: Effect of Drainage 1 : ATB ATB/DG AB 1
1 ' I . ’ '
1 Yes VS- N0 : : Controlling for other factors :
E 4” VS. 7” : : :
: 5 : 5
i E 1 Effect of Base Thickness i
: Effect of Base Type : : 12” vs. 16” l
i DGAB, ATB» ATB/DGAB, : E Controlling for other factors :
: PATB/DGAB, PATB/ATB : 0 g
: i : i
I : : :

 

Effects of design
features and site
factors

 

 

 

Figure 6-15 Methodology for site level analysis (SPS-l)

185

6. 5.1 Effects of design features on performance — Paired Comparisons at Level-A

A summary of p-values obtained from the non-parametric tests on RPIs (for all 18
sites) from level-A analyses is in Table 6-8. It is important to note that the signiﬁcance of
a factor indicates consistency in its effect across all the sites but not necessarily a

signiﬁcance of its effect on the magnitude of distress.

Table 6-8 Summary of p-values (non-parametric test) for Site Analysis - Level-A

 

 

 

 

 

 

 

 

 

 

Performance Measures
De51gn Factor 0:235:12 Rutting Roughness 33:11:: Loctgiztkliimal
WP NWP
HMA thickness 0.013 0.408 0.009 0.050 0.311 0.368
Base type 0.033 0.529 0.000 0.079 0.599 0.883
Base thickness - - - - - _ -
Drainage 0.047 0.056 0.020 0.040 0.035 0.028

 

 

 

 

 

 

The following is a summary of the main ﬁndings from paired comparisons at
level-A categorized by design factor and performance measures:

HMA Thickness

The effect of HMA thickness is consistent on roughness (IRI), fatigue cracking,
and transverse cracking, with 7-inch HMA pavements showing better performance. It is

not consistent for longitudinal cracking (WP and NWP)) and rutting (see Table 6-8).

0 Fatigue Cracking: HMA thickness appears to have a consistent effect on fatigue
cracking. Sections with 7-inch HMA thickness have consistently (across sites)
performed better than those with 4-inch HMA thickness. Twelve sites show a positive
effect, compared to ﬁve sites showing a negative effect and one site showing no

effect.

186

The effect of HMA thickness is less seen among the sites located in WF zone. Also, on
average, the superior performance of 7- over 4-inch HMA sections can be seen more for

sections on coarse-grained sub grade than for those on ﬁne-grained subgrade.

- Rutting: The effect of HMA thickness on rutting is not consistent. Nine sites show a
positive (lesser rutting in sections with 7-inch HMA surface thickness) effect and the
other nine show a negative effect, which shows no deﬁnitive trend of the effect across
sites.

- Roughness (IRI): The effect of HMA thickness on IRI is consistent across sites. On
average, sections with 7-inch HMA surface thickness have consistently performed
better than those with 4-inch HMA thickness. This trend was observed in thirteen out
of eighteen sites and two sites showed no effect.

0 Transverse Cracking: The effect of HMA thickness on transverse cracking is
consistent across sites. Sections with 7-inch HMA thickness have consistently
performed better than those with 4-inch HMA thickness. Eight out of eighteen sites
showed less transverse cracking for “7 inch” sections; seven sites showed no
transverse cracking, while three sites showed more transverse cracking for “7 inch”
HMA sections.

In terms of climatic zones, positive HMA thickness effect can be seen in all
zones except for WF. This could be attributed to the severe environmental conditions in

WF zone where even thicker HMA may not be able to inhibit cracking. Averaging over

all climatic zones, the HMA thickness effect is essentially the same for both subgrade

types.

187

Longitudinal Cracking-WP: The effect of HMA thickness on longitudinal
cracking-WP is not consistent. Nine sites show a positive effect and the other nine
sites show a negative effect, which shows no deﬁnitive trend of the effect across sites.
Longitudinal Cracking-NWP: HMA thickness seems to have no consistent effect
on longitudinal cracking-NWP. Nine sites show a positive effect and the eight sites

show a negative effect, indicating no deﬁnitive trend of the effect across sites.

Effect of Base Type

The effect of base type is consistent across sites on fatigue cracking, roughness, and

transverse cracking, with sections on DGAB showing the worst performance and sections

on ATB+PATB showing the best performance.

Fatigue cracking: Base type has a consistent effect, across sites, on fatigue cracking.
Sections with DGAB have shown the most amount of cracking while sections with
ATB+PATB have shown least amount of cracking. The order of performance from
best to worst is as follows: (1) PATB+ATB, (2) ATB, (3) ATB+DGAB (4)
PATB+DGAB and (5) DGAB.

Rutting: The effect of base type is not consistent. However, on average, sections with
PATB+ATB have slightly lesser rutting compared to sections with DGAB.
Roughness (IRI): The effect of base type on IRI is consistent across sites. Sections
with DGAB have consistently shown the highest IRI-values while sections with
ATB+PATB have shown the lowest values. The order of performance from best to
worst is as follows: (1) PATB+ATB, (2) ATB+DGAB, (3) ATB, (4) PATB+DGAB,

and (5) DGAB.

188

Transverse cracking: The effect of base type appears to be marginally consistent
across sites (p=0.079). Sections with PATB+ATB have somewhat lesser cracking
than sections with DGAB.

Longitudinal cracking-WP: The effect of base type is not consistent across sites.
Nonetheless, on average, the permeable bases appear to be performing better. The
worst performance was shown by the sections with DGAB.

Longitudinal cracking-NWP: The effect of base type is not consistent across sites.

However, on average, the best performing bases are the permeable bases-

PATB+ATB and PATB+DGAB, and the worst performing base is DGAB.

Effect of Drainage

The effect of drainage on all performance measures is consistent across sites, with

drained pavements showing better performance than un-drained pavements (see Table

6-8).

Fatigue cracking: Drainage has a consistent effect across sites on fatigue cracking.
Sections with drainage have performed better than those without drainage. Twelve
sites show a positive effect (better performance of drained sections), compared to ﬁve
sites that show a negative effect. Averaging over all climatic zones, the effect of
drainage can be seen better for sections with ﬁne-grained subgrade as opposed to
coarse-grained subgrade.

Rutting: The effect of drainage on rutting is consistent. Sections with drainage have
consistently performed better than those without drainage. Twelve sites show a
positive effect, compared to four sites showing a negative effect and two sites

showing no effect.

189

Roughness: Drainage has a consistent effect on roughness. Sections with drainage
have consistently (across sites) performed better than those without drainage. In
ﬁfteen out of eighteen sites drained sections have shown a better performance, while
reverse trend was found in only three sites. This effect is less seen among sections in
DF zone and could be attributed to the fact that drainage is not as important in this
zone.

Transverse cracking: Drainage has somewhat consistent effect on transverse
cracking. Sections with drainage have performed better than those without drainage in
most of the sites. Seven sites show a positive effect, while no transverse cracking
occurred in seven of the sites. Averaging over all climatic zones, the effect of
drainage is better seen for sections built on ﬁne-grained subgrade than for sections
built on coarse-grained subgrade.

Longitudinal cracking-WP: The effect of drainage on longitudinal cracking-WP is
consistent across sites. On average, sections with drainage have consistently
performed better than those without drainage. Eleven sites show a positive effect,
compared to three sites showing a negative effect and four sites showing no effect.
Moreover, the effect is more prominent for sections on ﬁne-grained subgrades as
opposed to those on coarse-grained subgrades.

Longitudinal cracking-NWP: Drainage has a consistent effect on longitudinal
cracking-NWP. On average, sections with drainage have consistently performed
better than those without drainage. Thirteen sites show a positive effect, compared to

three sites showing a negative effect and two sites showing no effect.

190

6. 5.2 Effects of design features — Paired Comparisons at Level-B

As explained in Chapter 4, level-B comparisons are more “controlled” compared
to level-A comparisons. To study the consistency of the effect of HMA thickness across
sites, nonparametric testing was performed on relative performance corresponding to 4-
inch and 7-inch HMA thicknesses, within each base type. Similarly, to investigate the
effect of base thickness, nonparametric testing was performed on relative performance
corresponding to 12-inch and l6-inch base thicknesses, within PATB/DGAB and
ATB/PATB. Also, nonparametric testing was performed on relative performance
correSponding to ATB and ATB/DGAB, within 8-inch and 12-inch base thicknesses. For
each of these effects, the corresponding p-values are presented in Table 6-9. A brief

summary of results from the level-B comparisons follows.

Table 6-9 Summary of p-values (non-parametric test) for Site Analysis - Level-B

 

 

 

 

 

 

 

 

Performance Measures
Design . Longitudinal
Factor 0:223) e Rutting Roughness Tdflartbsldfirse cracking
g g WP NWP
HMA 0.041 0.080 0.005 0.010 0.203 0.110
thickness
Base type 0.969 0.214 0.150 0.552 0.929 0.551
B?“ 0.307 0.022 0.046 0.933 0.499 0.387
th1ckness

 

 

 

 

 

 

 

HMA Thickness

The effect of HMA thickness can be examined for sections with different base
types. Among sections with DGAB, it was observed that HMA thickness has a positive
effect of on fatigue cracking, transverse cracking, and roughness (IRI). Also, on average,
a positive effect of HMA thickness was observed on sections with ATB for fatigue

cracking, but it is not consistent. The same effect was observed for rutting. This suggests

191

that increasing HMA thickness is more effective in pavement sections with unbound
bases as compared to those with treated bases.
Base Type

The effect of all ﬁve base types cannot be evaluated effectively for Level-B
comparisons since the only base types that can be compared are ATB and ATB+DGAB.
Nonetheless, the results show that the difference in performance between these two base
types is not statistically signiﬁcant.
Base Thickness

The effect of base thickness can be better seen when. using Level-B comparisons,
mainly because it is a more secondary effect relative to HMA layer thickness and base
type (treated versus untreated). Therefore it is more useful to look at this effect for two
types of (permeable) bases: (1)DGAB and (2) ATB.
The effect of base thickness was found to be consistent for rutting and roughness in
pavement sections with DGAB. The effect of base thickness is not consistent for fatigue
cracking and longitudinal cracking-NWP. However, on average, thicker (l6-inch) bases
have shown better performance than thinner (12-inch) bases. Finally, the effect of base

thickness is not consistent for transverse cracking and longitudinal cracking-WP.

192

6.6 OVERALL ANALYSIS

The results obtained from statistical analyses performed on the SPS-1 data are
presented in this section. Both the performance and response variables were analyzed to
study the effects of various design and site-factors on the pavement sections. Analyses
were performed combining all data and is referred to as ‘Overall’ analyses. Analyses
were also conducted in each climatic zone combining data from all sections within a zone
as per the recommendation of the project panel. Linear Discriminant Analysis (LDA),
Binary Logistic Regression (BLR), and Analysis of Variance (AN OVA) are the statistical
methods that were employed for analyses. Before presenting the results from statistical

analyses, the extent of distresses that occurred on the test sections is discussed.

6. 6.1 Extent of Distress by Experimental Factor

This section discusses the effect of the key experimental factors on performance
through the relationship between the magnitude and relative occurrence of the observed
distresses. Figures 6-16 through 6-21 show the percentage of test sections that have
exceeded various levels of distress for the key performance measures, categorized by
experimental (design and site) factors. Note that the effect of climatic zone is only shown
for the wet regions because of the limited number of sites in the dry regions (only four).

The following is a brief interpretation of these ﬁgures:

Fatigue cracking: Figure 6-16 indicates that about 70% of all test sections have shown
some fatigue cracking, with about 10% of all test sections showing 20% or higher

cracking by area. The effects of speciﬁc design and site factors are discussed below.

193

a)

b)

d)

HMA Thickness: About 75% of sections with thin HMA surface layer have
shown some fatigue cracking as compared to about 65% of sections with thick
HMA surface layer; the effect of HMA thickness tends to be larger for higher
levels of fatigue cracking.

Base Type: The difference in the percentage of test sections that have shown
fatigue cracking between those with unbound (DGAB) and those with treated
(ATB) bases is highest among all experimental factors (about 15%), with sections
built on DGAB bases showing the highest percentages.

Base Thickness: The effect of base thickness on fatigue cracking was found to be
insigniﬁcant.

Drainage: The effect of drainage in terms of higher percentage of test sections
showing fatigue cracking is more pronounced at the lower levels of fatigue; the
effect becomes insigniﬁcant at the later stages of fatigue. This could mean that
drainage is more effective in the early life of the pavement, and becomes less
effective later in the pavement life. Also, the effect of drainage is slightly more
visible for ﬁne-grained than for coarse-grained subgrade soils [Figure 6-16 (d)].
Climatic Zone: There are consistently more sections in wet-freeze (WF) than wet-
no-freeze (WNF) climate that have shown fatigue cracking exceeding various
levels, with about 10% more sections in WF than in WNF climate.

Subgrade Type: There are consistently about 15% more sections built on ﬁne-
grained than coarse-grained subgrade soils that have shown fatigue cracking
exceeding various levels, and the effect of subgrade soils tends to be larger for

higher levels of fatigue cracking.

194

Rutting: Figure 6-17 indicates that about 60% of all test sections have shown rut depths

higher than 0.25 inch (6.25 mm), and about 20% of all test sections showing rut depths

higher than 0.5 inch (12.5 mm). The effects of speciﬁc design and site factors are

discussed below.

a)

b)

d)

HMA Thickness: The effect of HMA thickness on rutting was found to be
negligible.

Base Type: There are about 10% to 15% more sections with unbound (DGAB)
bases that have rut depths greater than 7.5 mm than those with treated (ATB)
bases. This difference is relatively constant at higher rut depths.

Base Thickness: There is a slight effect of base thickness for sections that have rut
depths that are less than 7.5 mm, with about 5% more sections with thinner (8
inch) bases than those with thicker (16 inch) bases. The effect becomes less
apparent for rut depths greater than 7.5 mm.

Drainage: There are consistently about 5% more seetions without drainage than
with drainage that have exceeded various rut depth levels. Also, the effect of
drainage is slightly more visible at the higher rut depths and for ﬁne-grained
subgrade soils [Figure 6-17 (d)].

Climatic zone: The effect of climatic zones (within wet regions) on rutting
appears to be more signiﬁcant at rut depth higher than 7.5 mm, with about 10%
more sections in wet-freeze than in wet-no-freeze climate exceeding various rut
depths.

Subgrade Type: There are consistently about 10% more sections built on ﬁne-

grained than coarse-grained subgrade soils that exceed various rut depths.

195

Roughness: Figure 6-18 indicates that about 60% of all test sections have shown IRI

values higher than 1 m/km, with about 20% of all test sections showing IRI values higher

than 1.4 m/km. The effects of speciﬁc design and site factors are discussed below.

a)

b)

d)

HMA Thickness: The percentage of test sections with thin (4 inch) HMA surface
layer that have exceeded an IRI of 1.2 m/km is about 40% as compared to about
20% for test sections with thick (7 inch) HMA surface layer. The percentage of
test sections with thin (4 inch) HMA surface layer exceeding higher IRI levels is
5% to 10% more than that of test sections with thick (7 inch) HMA surface layer.
Base Type: The percentage of test sections with unbound aggregate base (DGAB)
that have exceeded an IRI of 1.2 m/km is about 40% as compared to about 20%
for test sections with asphalt treated base (ATB). The percentage of test sections
with a 8-inch base exceeding higher IRI levels isiabout 10% to 15% more than
that of test sections with an l6-inch base.

Base Thickness: The effect of base thickness on roughness is more pronounced
than for other performance measures, showing a percentage of test sections with a
DGAB base exceeding 1.2 m/km and higher IRI levels that is about 10% to 15%
more than that of test sections with an ATB base.

Drainage: There are consistently about 5% more sections without drainage than
with drainage that have exceeded various IRI levels.

Climatic zone: The effect of climatic zones (within wet regions) on roughness
appears to be the most signiﬁcant, with about 20% to 30% more sections in wet-

freeze than in wet-no-freeze climate exceeding 1.2m/km and higher IRI levels.

196

t) Subgrade Type: The effect of subgrade type on roughness is more pronounced
than for other performance measures, with about 15% to 30% more sections on
ﬁne-grained than on coarse-grained subgrade exceeding 1.2m/krn and higher IRI

levels.

Transverse cracking: Figure 6-19 indicates that about 40% of all test sections have shown
some transverse cracking, with about 10% of all test sections showing 20m or higher
length of transverse cracking. The effects of speciﬁc design and site factors are discussed
below.

a) HMA Thickness: Only a slight effect of HMA thickness was found on transverse
cracking.

b) Base Type: Base type appears to be a signiﬁcant factor affecting transverse
cracking. About 10% to 15% more test sections with unbound aggregate base
(DGAB) than those built with an asphalt treated base (ATB) at various levels of
transverse cracking.

c) Base Thickness: Only a slight effect of base thickness was observed.

(1) Drainage: Only a slight effect of drainage on transverse cracking was observed.

e) Climatic Zone: Climate seems to be a signiﬁcant factor affecting transverse
cracking. There are about 15% to 20% more test sections in WF zone than those
built in WNF zone at various levels of transverse cracking.

f) Subgrade Type: Subgrade soil type seems to have some effect on transverse
cracking, in that, slightly higher proportion of sections built on ﬁne-grained soils

have shown cracking compared to those built on coarse-grained soil.

197

Longitudinal cracking: Figures 6-20 and 6-21 indicate that about 50% of all test sections

have shown some longitudinal cracking-WP and about 75% of all test sections have

shown some longitudinal cracking-NWP. The effects of experimental factors are

discussed below.

a)

b)

d)

HMA thickness: HMA thickness appears have a negligible effect on longitudinal
cracking.

Base Type: There seems to be a slight effect of base type on longitudinal
cracking, in that, sections with ATB has shown lesser cracking than those with
DGAB.

Base Thickness: Only a slight effect of base thickness was observed. Sections
with l6-inch base thickness have slightly less occurrence of cracking than those
with 8-inch base.

Drainage: There appears to be some positive effect of drainage on lower levels of
longitudinal cracking-WP. However this effect was observed to be negligible for
higher levels of cracking.

Climatic Zone: The effect of climatic zone (within wet regions) on longitudinal
cracking is more pronounced than other effects, especially for longitudinal
cracking-NWP. About 10% to 20% more test sections in wet-freeze than in wet-
no-freeze climate exceed 100 m or more of longitudinal cracking-WP, and about
20% to 35% more test sections in wet-freeze than in wet-no-freeze climate exceed
100 m or more of longitudinal cracking-NWP.

Subgrade Type: Only a slight positive effect of subgrade type was observed for

longitudinal cracking-WP.

198

o\onm

wagons osmsﬂ co 280$ 3808.3de mo 80am £6 oSmE

 

 

 

 

 

$8

 

 

090 03.35 S
2.6.8.0 2&3 .x.
e\oom .32 ..\oo— e\ow ...\..o
A _ a .0 ﬁg
{com
.596
. $8
.
2.... it - ea...
AXbo—
omeBQ on
”...an 2&8“. .x.
$8 $2 $2 .xem e\oo
2.6
$8
$3
_ $8

 

 

- 08....

suouoas 189130 %

suorioas 189110 %

anon owns—:0 E 39.33 nocmﬁwiocc 53> omeED AB

 

 

     

 

 

 

 

 

 

 

  

 
 

  

mic—8.5 ozwzam .x. weﬁoﬁu 2.96.... .x.
2mm 28 $2 2... 0% exec 58 $8 $2 exec. exen ..er
. . _ . ..er $0
%
. $8 6 $8
.19.
1 5.... m
.00. $0..
. .58 m
“.23 101 1. <68 m omnEED um: .58
"S, 161 emanate oZ 101 .
e\eoo. 1 exec»
65 68m SD 0.85.2... <2... 3
mﬁxoﬁu ozwzam e\e wee—8.0 Swan... ..\e
$2 $8 exem. 2... 08m x... 08mm .88 0%. 08... ea 08..
J i2 i. . {a 1. iul « u G\&
%
$8 6 $8
I:
089. m 2....
m<OD+mF< I01 w .38 m $8
0.
m...< IT 0. $8 w $8
950 101
$8. 1 e88.

 

 

 

 

 

 

suouoas 19:91 go 0/0

suouoas 1591 }O 1y0

199

odb otﬁwnam 5

25.5893.
m. 2. o. 2 .0. ma

 

d

3.230 I01

”numb IOI-

 

 

 

285.3... 8.5 A8

as... and 33.
m. 3. o. 2 m 3

 

_ d 1 _l _ , 4

 

 

maﬁa co 280$ .Scoﬁtodxo mo Scam :10 oSwE

o\oo

..\oo~

$9»

$8

o\com

.39:

..\co

o\oom

$9.

o\ooo

.Voow

o\ooc_

0:8. 33.55 E

 

 

 

 

 

Es. sad .3.
n. mN— S n6 n ON
%
m.
m
w...
m.
U
5 “~23 191
an; l¢l
090 Sam. 3v
3:5 sad 23.
2 Wm. 9 ms m ON

%
m.
m
S
m. 9.093... 11
m E< 101

m<00 l

 

 

 

o\eo

Qoom
o\aov
$8
$8

$09

o\oo
A.\aom
$9.
$8
{cox
o\ooo _

SUOIIOOS 183) JO 0/0

suouoas 1591 go %

unnamed nocﬁcwiocc 5:3 emacﬁku A3

2:5 580 .3.
m. Wm. o. m... m 0m 0

i9114 .1 q r .—

QWNCEHD liml
omeEQ oZ 1?.

Iii Ill ll ill 111|lrl|11111 liliL

 

 

r

 

 

85.2.: <2: a.

.85 £80 .3.

2 m.N~ 2 m8 m mN o

 

_ u A —

 

 

o\oo

.xbm

.. $9V

{coo

{new

a\ooo _

o\oo

- o\aoN

...\e9V

$00

1 .xbw

. execs.

suouoas 1291p %

suouoas 1591 30 %

200

 

 

 

 

 

 

 

 

cab unﬁwnzm 8
93.52.
m. 9. a... N. _ md ed
L
8800+ L
vcmmlol
$053.5 8mm Gv
AEéEvE.
M... c.— v. N. . ad ad

 

 

$05.98. .8 390mm 3:08.598 mo Hoobm 3-0 85me

go

o\eom
gov
oboe
$3

$8 _

e\oo
$8
$9.
$8
foo»
o\ooo _

25a 0.38:0 A8

 

 

 

  

 

 

 

 

 
 

       

AEEVE.
m M... o. v. m. . ...o 0...
ill!» Ilillll_..u||.llll. 1 q 4 °\°o
% L {com
0
7.0..
w 1 $8
m. 38
U
s ”.321? 4 $8
SIOI
- r n$8.
093 8.5 SV
.52....
N w. c. 3 .2 _ ...o o...
q 4 . . O\Oo
% 1 $8
0
1..
m 1 $8
8
m. 38+E<+ 1 $8
0
w MET? L $9.
m<ODIOI
£8.

 

 

 

suopoos 19150 %

omeSD AB

.52.. a.

w. o. v; N._ _ wd

ed

 

 

«

    

d

 

A.

 
 

owncﬁﬁ ImT
09.580 02 .4:

 

 

suopoas 1591 go %

m35.2... <2: 3

Act—>5 3:
w. o. v; N. _ wd

 

 

o\oo

gem
o\eov
goo
30w

$00.

ed

 

 

 

o\oo

A.\..o~
{new
$8
$3

$8—

suopoas 1831p %

SUOUOQS 1891 }0 %

201

 

weigh. 383...... so €82... geogomxo mo Hootm 3-0 8:me

 

 

 

 

 

 

 

25 2.835 9
c5 9.3080 8.26:8...
nm em m. o. o
d A u I <8
- $3
_ o\ocv
. $8
8:80 1? . $8
05...— lol
$.8—
mmoqon. 8mm A8
A8. wﬁonU 8.535....
mm cm .2 o. o
q . $o
. $8
0~ IO.-
N. .o. . $8
E...
l $8.

 

 

suogaos muo %

8110930510 %

anon 288:0 3

$5 95.080 8535......

 

L231? ..
$19..

 

 

 

090 8mm 3V

AEV 98.080 3.055....

mN cm 3 o. m o

 

11‘. A ‘

 

a8é< l...
m...< 1T .

e\ecw

AXbo—

o\oo

gem

$9.»

«80

«8a

 

$819..

 

f till I...

.38—

suopoos 191 30 %

suopoos JO %

mm cm

09.595 .3

$5 wcﬁoﬂu 8855......

m— o. m o
o\oo

 

d

 

””0555 I?
ommcmﬂo OZ 1T

. — a

4 $8
" $9.
. $8

. c\ocw

 

o\ooo_

 

mm om 2 o. n

305.2... <3... 3

$5 wet—8.0 883.52...

C

o\ao

 

4

5+

#10!

 

a d d

- .XbN
" $3
- {ace

4 {new

 

o\aoo_

 

5009398 1891 30 %

suogaos 19130 %

202

$588880 38838.5. ..o 8808 8% mo Scam omé 23w...

 

 

 

 

 

 

 

 

0...». 8835 C.
.E. .3835 35383
o8 8m 8. 8. on o
u.- q 4 . {5
0300 Iﬂl J {60%
0:.... IT
1 $8.
$05.25 08m A...
.E. 8.88.86 25.5.83
..nN 8.... on. 8. cm o
- 1 . $..
. {can
/. 8......
.
-. o
o. no. \uoe
N. no: . $3
w I9:
$8.

 

 

suopaos 19150 %

 

 

econ 0325.0 9.

.s. 8.88.8.0 35.5.83

 

 

 

 

Ill — 4 «

suopozs an: 30%

 

  

..
.8
$8
w .xbv
- $8
L23 |¢I . {com
um; IOI
$8.
25 8mm 3.
.5. 8.95.8... .2383
..8 ..8 a. 8.. ..m o
$..

 

 

suopoos 39150 %

5009305 1801 }O %

 

 

 

 

 

  

 

em.........._n.....
.5. .38...qu 35.5.83
8.. 08 ..8 8. 8. 8 ..
ﬂ . u . °\Oo
, $8
.$8
.$8
w
o 2851.. . $8
03580021...
$8.
“35...... <5... .n.
.6. 8.83020 3:62.83
.8 8m 8. 8. 8 o
. . 1.... $..
- $8
$9.
.
- $8
To. . $8
VIOI
1 .l..!...[ iii--- $8.

 

 

suogoos 1991 30 %

suopoos m: 30 %

203

352883080 188.85me.. :0 £808.. 28 mo Seam 3.0 onE

 

 

 

 

 

 

 

2.3 3.8825 G.
.5. 8.2.88.5 853.85.
o8 88 8. 8. 8 o
.— ﬁ 4 4 °\&

1 $8

$6..

. $8

. $8

05m lol
?, 1.. :L O\68—
82.on. 08m 6.
.5. 8.2.8.8020 883.83
O8 88 8. 8. 8 o

. . . . $..
1 $8
, $8
8. I? 1 $8

.
N. I? ,. $8
w Io:
[ in $8.

 

 

 

 

suopoos m; _;o %

51109309159130 %

econ 28.5.0 8.

 

 

 

 

 

 

.5. 832.9885 .8383
on 8 8. 8. 8 o
u . . .\oo
1 $8
1 $8
. $8
823 1.. L. $8
"5’ IOI
$8.
25 08m 3.
8.. 832.8585 883.83
o8 com on. 8. on o
a J 4 4 {a
1 $8
I $9~
.. e
m<8+mh< IOI \uoe
PE... .. $8
$8 Ibl
$8.

 

 

 

suogaoas 180) JO %

suopoos 19m JO %

82.80 8.

8.. “5.2.8285 85.5.85.
88 8. 8. 8 o

-, . . 111i; .8

0mm

 

..\a0m
, $9.

828.0 no. . $8

$5.85 02+

 

 

$.8—

 

m35....... <2: 8.

.5. 8.2.8886 3:83.83

o\oo
‘ {com
- €99.
, $8

.
a $8

 

 

 

$8.

SUODDGS 153) JO "/0

suopaos Isa: go %

204

6. 6.2 F requency-based methods

Two frequency-based methods were used- Linear Discriminant Analysis and
Binary Logistic Regression (details in Chapter 4). The results from these analyses are as

follows:

Discriminant Analysis
In this analysis, two mutually exclusive groups were deﬁned as follows:

0 Alligator, transverse and longitudinal cracking: Cracked versus non-cracked.
o Rutting: Rut depth < 7 mm versus rut depth > 7mm
- Roughness: IRI < 1.4 m/km versus IRI>1.4 m/km

This analysis was intended to identify the experimental factors which help in
discriminating the cracked versus non-cracked pavement sections. As most of the
pavements in the SPS-1 experiment have not shown a high level of distress, this analysis
will help in ﬁnding the signiﬁcant design and site factors contributing to occurrence of
distress. In order to include the effect of trafﬁc and pavement age, these were considered
as covariate in this analysis. Table 6-10 summarizes the results of network level analysis.

The performance measures were deﬁned as dichotomous variables and all design
and site factors were used as independent variables. The following summarizes the
results from this analysis:

' 0 Fatigue cracking: The effects of drainage condition and base type were found to
discriminate between cracked and non-cracked sections. Test sections without
drainage built on unbound (DGAB) bases are more likely to crack.

' Ruttz’ng: The effects of drainage condition, subgrade soil, base thickness were found

to discriminate between sections having rut depths greater or less than 7mm. Test

205

sections without drainage with thinner bases and built on ﬁne-grained subgrade soils
in wet zones are more likely to exhibit severe rutting.

o Roughness: The effects of climatic zone, subgrade soil and base thickness were
found to discriminate between sections having IRI greater or less than 1.4 m/km.
Test sections with thinner bases built on ﬁne-grained subgrade soil and in wet freeze
zone are more likely to exhibit higher roughness. Sections with higher initial
roughness are more likely to become rougher with age.

0 Transverse cracking: The effect of base type to a lesser degree was found to
discriminate between cracked and non-cracked sections. Test sections in wet freeze
zone with unbound (DGAB) bases are more likely to crack. Also, older sections are
more likely to crack.

0 Longitudinal cracking: The effect of climatic zone was found to discriminate

between cracked and non-cracked sections (inside the wheel path). Test sections
built in wet-freeze zone are more likely to crack outside the wheel path. Also, older

sections are more likely to crack (in and outside the wheel path).

Table 6-10 Summary of p-values from LDA for determining the effect of experimental
factors on pavement performance measures

 

 

 

 

 

 

 

 

 

 

Performance Measures
Design Factor Fatigue - Transverse Longitudinal
cracking Rutting Roughness cracking cracking

WP NWP
HMA thickness 0.39 1.000 0.370 0.320 0.88 0.310
Base type 0.098 0.517 0.250 0.139 0.77 0.690
Base thickness 0.92 0.077 0.076 0.736 0.19 0.421
Drainage 0.09 0.056 0.370 1.000 0.47 0.310
Subgrade type 0.045 0.011 0.000 0.177 0.184 0.001
Climatic Zone 0.392 0.578 0.000 0.417 0.002 1.000

 

 

 

 

 

 

 

 

206

Logistic Regression

The binary logistic regression model was used to model the probability of
occurrence for the various performance measures. This method requires fewer
assumptions than discriminant analysis and even when the assumptions required for
discriminant analysis are not satisﬁed, it performs well. The overall models for each of
the performance measures were found to be signiﬁcant. The results using the maximum

likelihood method are summarized in Tables 6-11 and 6-12.

0 Fatigue cracking:
HMA Thickness— Thin (4-inch) pavement sections have a slightly higher
probability of cracking than thick (77inch) sections when all other variables are
held constant.
Base Type— Pavement sections with unbound (DGAB) base have a signiﬁcantly
higher probability of cracking than those with bound (ATB/DGAB) bases.
Drainage— Pavement sections with no drainage have a higher probability of .
cracking than those sections with drainage.
Climatic Zone— Pavement sections in freeze zones have a signiﬁcantly higher

probability of cracking than those in the no-freeze environments.

0 Rutting:
Drainage— Pavement sections with no drainage have a slightly higher probability

of rutting (rut depth > 7 mm) than those sections with drainage.

207

Subgrade Type— Pavement sections built on ﬁne-grained subgrade soils have a
signiﬁcantly higher probability of rutting than those sections built on coarse-
grained subgrade soils.

Climatic Zone— Pavement sections in WNF zones have a signiﬁcantly higher

probability of rutting than those in the WP environments.

0 Roughness:
HMA Thickness— Thin (4 inch) pavement sections have a higher probability of
showing higher roughness (IRI > 1.4 m/km) than thick (7 inch) sections.
Base Type— Pavement sections with unbound (DGAB) base have a signiﬁcantly
higher probability of showing higher roughness than those with bound
(ATB/DGAB) bases.
Base Thickness— Pavement sectionswith thin bases have a signiﬁcantly higher
probability of showing higher roughness than those with thick bases.
Subgrade Type— Pavement sections built on ﬁne-grained subgrade soils have a
signiﬁcantly higher probability of showing higher roughness than those sections
built on coarse-grained subgrade soils.
Climatic Zone— Pavement sections built in wet-freeze zone have a signiﬁcantly
higher probability of showing higher roughness than those sections built in wet-

no-freeze zone.

208

Table 6-11 Summary of p-values from BLR for determining the effect of experimental
factors on pavement performance measures (Wet zones)

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

Performance Measures
. , Longitudinal
D t
681811 Fac or £225: Rutting Roughness 23:51:; se cracking
g g wp NWP
. 0.160 0.833 0.068 0.493 0.360 0.31
HMA “mm“ (1.8) (1 . 1) (3.7) (0.6) (1.5) (0.7)
Base 0.024 0.972 0.006 0.711 0.437 0.396
type (2.4) (1.0) (33) (1.2) (1.7) (1.7)
Base thickness 0.420 0.212 0.038 0.632 0.410 0.733
(1.7) (2.5) (14) (1.3) (1.8) (0.8)
Draina c 0.045 0.124 0.278 0.316 0.40 0.837
g (2.8) (2.2) (2.4) (2.7) (0.6) (0.9)
Sub ade type 0.960 0.015 0.000 0.345 0.000 0.009
3‘ (1.0) (3.4L (571) (0.005) (22) @34)
. . 0.088 0.098 0.000 0.976 0.73
Climatic Zone (22) (.42) 9129 0.316 (10) (1.0) (0.862)

 

 

 

Note: The values in parenthesis are odds ratios

Table 6-12 Summary of p-values from BLR for determining the effect of experimental
factors on pavemengierformance measures (All zones)

 

 

 

 

 

 

 

 

 

 

Performance Measures
Design . Longitudinal

Factor 3:25:12 . Rutting Roughness T335123: cracking_
wp NWP

HMA 0.527 0.34
thickness 0.19 (1.5) 0.98 (1.0) 0.068 (3.7) 0.224 (0.5) (125) (0.7)
Base type 0.013 (2.2) 0.98 (1.0) 0.006 (33) 0.043 (3.2) 8545) (02202)
Base 0.376 0.77
thickness 0.81(0.8) 0.4 (2.3) 0.038 (14) 0.974 (1 .0) (1. 6) (1.2)
Drainage 0.009 (3.1) 0.22 (1.7) 0.278 (2.4) 0.473 (1.6) (2173;) (£122;
Subgrade ’ 0.000 0.076 0.019
we 0.073 (0.5) 0.87 (1.1) £571) 0.006 (0.001) (2.2) (0.42)
0.001 0.019 0.002
Climatic 0.000 0.747 - - 0.254 0.003
Zone 0 000 (12) 0.72 - - _ (5.1)
WF-DNF 0' 002 (9) (0.83) - - _ 0.003
WNF-DNF 0 000 0.611(1.4) (420) (17) (6.4)

. (23) -

DF-DNF (1 4) (0 76) 0.000
WF-WNF ' (0.6) ' (27)
(0.8)

 

 

 

 

 

 

 

 

Note: Very high value of odds ratio is caused by the un-balanced data or too few sections in one of the
categories.

209

0 Transverse cracking:
Climatic Zone— Pavement sections built in wet-freeze zone have a higher
probability of cracking than those sections built in wet-no-ﬁ'eeze zone.

Also, older pavement sections have a signiﬁcantly higher probability of cracking.

0 Longitudinal cracking:
Subgrade Type— Pavement sections built on ﬁne-grained subgrade soils have a
higher probability of longitudinal cracking in the wheel path than those sections
built on coarse-grained subgrade soils.
Climatic Zone— Pavement sections built in freeze zone have a signiﬁcantly
higher probability of cracking (outside the wheel path) than those sections built in
no-freeze zone.

Also, older pavement sections have a signiﬁcantly higher probability of cracking.

210

6. 6.3 Analysis of Variance

Several analyses of variance (ANOVA) were conducted for each of the
performance measures and response indicators. The first AN OVA was targeted at
determining the signiﬁcanCe of only the main structural design factors considered in the
experiment. This was achieved by blocking the site factor (to neutralize the effects of
subgrade type, climatic conditions, trafﬁc, age, and construction variability) as well as
accounting for the variability in target layer thicknesses. The main structural design
factors included in the ANOVA are listed below:

HMA thickness (4-inch versus 7-inch)

Base type (DGAB, ATB or DGAB+ATB)

Base thickness (8 inch, 12 inch or 16 inch)

Drainage condition (with versus without Permeable ATB)

To meet the assumptions of ANOVA, the dependent variables (performance
measures) had to be transformed using the natural logarithm. This was particularly
relevant for all cracking distresses because of the large number of zeroes in those
populations. A negative consequence from this is that the number of sections used in the

analysis is reduced.

6.6.3.1 Effect of Design Factors on Pavement Performance
The results from this analysis are summarized in Table 5-13 and indicate that the
most signiﬁcant design factor is the base type, which has a signiﬁcant effect, statistically
as well as operationally, on all performance measures. The AIRI (which is the change in
IRI. between initial and latest value) is also signiﬁcantly affected by base thickness. The

initial roughness is signiﬁcantly affected by all the design factors except for drainage.

211

Also, the effect of drainage condition on rutting, and the effect of base thickness on
longitudinal cracking-NWP and change in roughness are statistically signiﬁcant.

For investigating the mean difference between the levels of design factors, the
marginal means (predicted cell means from the model) were transformed back to the
original scale of the distress. These conversions were necessary in order to ﬁnd out the
practical/operational mean difference. The marginal means were back transformed using
the properties of lognormal distribution. A random variable X is considered to have a
lognormal distribution if Y=ln (X) has a normal probability distribution, where 1n (X) is
the natural logarithm to the base e. Equations (5-1) and (5-2) are used to calculate the

mean and variance of a random variable X.

1 2
.ux : exp(,uy "FED-y j (5-1)

2 2 2
ox :ﬂx [exp(0'y )—1] (5-2)

Where: 11, and 6,2 are the mean and the variance of lognormal distribution.

The marginal means of performance measures (for which natural logarithmic
transformation was necessary to meet the ANOVA assumptions) were estimated by using
equation 5-1. The mean squared error (MSE) was considered as the “best” estimate of
the variance for lognormal distribution in all analyses. Table 6-14 shows the back
transformed marginal means for all levels of design factors in the SPS-1 Experiment. The

following discussion summarizes the effect of key design factors on performance:

212

0 Effect of base type: The effect of base type was found to be signiﬁcant for all
performance measures except for rutting. Pavement sections with dense-graded
aggregate bases (DGAB) have shown the worst performance for all distresses while
those with asphalt treated bases (ATB) have shown the best performance. Sections
built with DGAB have shown signiﬁcantly (operationally and statistically) higher
fatigue cracking compared to those built with ATB. On average higher rutting was
observed on pavement sections built with DGAB than those constructed with ATB. '
In the case of other distresses (change in roughness, transverse cracking, and
longitudinal cracking) the difference in the performance of sections built on DGAB
and sections built on ATB were only found to be statistically signiﬁcant (i.e., they

are not of operational signiﬁcance at this point in time).

0 Effect of HMA thickness: In general, thin [4-inch (102 mm)] pavement sections
were built rougher than thick [7—inch (178 mm)] pavements. On an average, thin
pavements [4-inch (102 mm)] have shown slightly higher fatigue cracking and
rutting than thick [7-inch (178 mm)] pavements. However, this effect was found to

be of marginal statistical signiﬁcant.

0 Effect of base thickness: Sections with thicker bases [12—inch (305 mm) and 16-
inch (406 mm)] were built smoother compared to those with thinner base [8-inch
(203 mm)]. Also, sections with thinner base [8-inch (203 mm)] have shown more
change in roughness than those with thicker base [12—inch (305 mm) and 16-inch

(406 mm)]. However, this change in roughness is not of practical signiﬁcance. On an

213

average, pavement sections with 8-inch (203 mm) base have shown more rut depths
than those with 12-inch (305 mm) and 16-inch (406 mm) thick bases. However, this
effect is not of practical signiﬁcance. More longitudinal cracking-NWP occurred in
sections built with 8-inch (203 mm) base compared to those with 16-inch (406 mm)

base.

Effect of drainage condition: Onaverage, pavement sections with drainage have
shown slightly lower rutting than those without drainage; however this difference in

performance is not statistically signiﬁcant.

214

Table 6-13 Summary of p-values from ANOVA for determining the effect of main design

actors on pavement performance measures—Overall

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

Performance Measures '
D ' . Longitudinal
13:3: Fatilgue le . IRI Transgerse crackin
crac mg depth AIRI IRIO crac mg WP ' NL—WP
HMA 0.163 0.074 0.870 0.006 0.758 0.737 0.787
Thickness
1r
Base Type 0000 0.510 0.004 0.000 0.016 0.079 0.031
Base 6
Thickness 0.951 0.080 0.027 0.028 0.697 0.488 0,008
Drainage 0.347 0.250 0.293 0.160 0.544 0.645 0.874
Site
3 locked) 0.000 0.000 0.000 0.000 0.000 0.000 0.000
R2=0.712 R2=0.55 R2=0.624 R2=0.603 RT=0630 R2=0.72 R2=0.770
N=133 N=159 N=163 N=212 N=75 N=97 N=140

 

 

 

Ifote: The model considered for this analysis only has main effects for al

Also shows operational/practical signiﬁcance,

Structural rutting only

 

design factors.

Table 6—14 Summary of marginal means from AN OVA for determining the effect of

Design Factor

HMA
Thickness

Base Type

Base
Thickness

Drainage

MSE

N

Y

(1.51)

are

(0.06)

ALRI IRIo
(In/km)

45

215

(in/km)
0.837
0.786

0.87

0.79

Transverse
cracking

(m)

1.4

(1.3)

WP
(m)

(1.4)

 

(0.87)

The AN OVA was conducted for the design factors within each climatic zone as per the
project panel recommendations. However, this analysis suffers from the lack of data
within zones, especially within “Dry” zones, where only 2 sites each are available for DF
and DNF zones. The results of ANOVA for “Wet” zones are more reliable as 8 and 6
sites are available within WP and WNF zones, respectively. The following discussion
summarizes the effect of key design factors on performance in WF climatic zone (see
Table 6-15 and Table 6-16):
0 Effect of HMA thickness: In general, thin (4-inch) pavement sections were built
rougher than thick (7-inch) pavements. On average, thin pavements (4-inch) have
shown slightly more fatigue cracking and rutting than thick (7-inch) pavements.

However, this effect was not found to be statistically signiﬁcant.

0 Effect of base type: The effect of base type was found to be signiﬁcant for all
performance measures except for transverse and longitudinal cracking. In the case of
distresses that are signiﬁcantly affected by base type, pavement sections with dense-
graded aggregate bases (DGAB) have shown the worst performance while those with
asphalt treated bases (ATB) have shown the best performance. Sections built with
DGAB have shown signiﬁcantly (operationally and statistically) more fatigue

cracking, rutting, and change in roughness compared to those built with ATB.

0 Effect of base thickness: Sections with 12-inch (3 05 mm) bases were built smoother
compared to those with 8-inch (203 mm) base. Also, sections with 8-inch (203 mm)

base have shown signiﬁcantly (practically and statistically) higher change in

216

roughness than those with 16-inch base. However, the change in roughness was not
found to be practically signiﬁcant between sections with 8-inch (406 mm) base and
those with 12-inch (305 mm) base. On average, pavement sections with 8-inch (203
mm) base have shown more rut depth than those with 12-inch (305 mm) and 16-inch
(406 mm) thick bases. However, this effect wasnot found to be statistically
signiﬁcant. More longitudinal cracking-NWP occurred in sections built with 8-inch

base compared to those with 16-inch base.

Effect of drainage condition: Pavement sections with drainage have shown less
rutting than those without drainage. This difference in performance was found to be

both statistically and practically signiﬁcant.

217

Table 6-15 Summary of p-values from AN OVA for determining the effect of design
factors on ﬂexible pavement performance—WP Zone

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

Performance Measures
. , Longitudinal

D9518“ Factor Fatigue Rut 1R1 Transverse cracking

cracking Depth AIRI IRIO crackmg WP NWP
1! I. 0.745 0.688 0.277 0.133 0.560 0.893 0.762
th1ckness

* i *

Base type 0004 0,001 0,075 0.012 0.128 0.232 0.400
Base thickness 0.832 0.504 0040‘ 0.084 0.278 0.873 0,059'
Drainage 0.674 (1,012" 0.874 (1003" 0.359 0.813 0.885
Site (blocked) 0.000 0.000 0.000 0.000 0.000 0.000 0.000

R2=0.638 R2=0.631 R2=0.620 R2=0.628 R2=0.71 R2=0.695 R2=0.813

N=58 N=92 N=76 N=92 N=3 l N=36 N=60

 

 

 

up

Also shows operational/practical signiﬁcance

Table 6-16 Summary of marginal means from ANOVA for determining the effect of

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

main desi n factors on pavement performance measures— WF Zone
Average Performance
F t' R t IRI T Longitudinal
“swam c.2355; aeﬁa 53251313? "““mg
(sq-m) (m) “R1 “‘10 (m) WP NWP
(m/km) (m/km) (m) (m)
4,, 17.6 6.1 0.50 0 891 8.7 24.5 29.3
HMA (2.3) (1.76) (-0773) ' (1.92) (2.26) (3.22)
Thickness 7,, 15.9 5.9 0.55 0 844 7.4 23.1 31.75
(2.2) (1.73) (-0.673 ' (1.76) (2.2) (3.3)
34.4 7.5 0.55 11.0 23.1 26.5
DGAB (3.0) (1.96) (0665) 0932 (2.15) (2.2) (3.12)
10.3 5.7 0.45 5.7 11.5 33.4
835° Type ”B (1.8) (1.68) (0865) 0333 (1.5) 41.5) (3.35)
14.0 5.2 0.56 8.3 47.5 30.2
ATB/DGAB (2.1) (1.6) (-0.64) 0'84 (1.87) (2.92) (3.25)
8,, 17.1 6.4 0.61 0 91 11.5 28.5 38.8
(2.3) (1.8) (-0.561) ‘ (2@ (2.@ (3.5)
Base 12,, 18.8 5.8 0.55 0 833 8.3 23.6 33.0
Th1ckness (2.4) (1.7) (-0.673) ' (1.87) (2.22) (3.34)
16,, 15.4 6.0 0.42 0 86 5.7 19.7 21.3
(2.2) (1.74) (-094) ' (1.49) (2.04) (2.9)
N 18.1 6.8 0.53 0 924 9.4 22 30.5
Drainage (2.4) (1.86) (-0714) ' (2.0) (2.15) (3.26)
Y 15.4 5.4 0.52 0 811 6.9 25.5 29.6
(2.2) (1.63) (.0732) ‘ (1.69) (2.3) (3.23)
MSE (1.073) (0.113) (0.139) (0.493) (1.881) (0.316)

 

 

 

 

 

 

 

 

 

Note: Values in parenthesis are the lognormal marginal mean values.

218

 

The following discussion summarizes the effect of key design factors on performance in
WNF climatic zone (see Table 6—17 and Table 6-18):

0 Effect of base type: The effect of base type was found to be signiﬁcant only for the
change in roughness. Pavement sections with dense-graded aggregate bases (DGAB)
have shown higher change in roughness compared to those with asphalt treated bases
(ATB). However, difference was not practically signiﬁcant.

0 Effect of HMA thickness: On average, thin pavements (4-inch) have shown slightly
more fatigue cracking compared to thick (7-inch) pavements. This effect was found
to be both statistically and practically signiﬁcant.

0 Effect of base thickness: The effect of base thickness on various performance
measures was found to be statistically insigniﬁcant. However, on average, higher
change in roughness was observed in sections with 8-inch base compared to those
with 12-inch or 16—inch base.

0 Effect of drainage condition: Pavement sections with drainage have shown less
rutting than those without drainage. This difference in performance was found to be

both statistically and practically signiﬁcant.

The fractional factorial design for the SPS-1 experiment calls for a tradeoff
between selecting the number of “runs” and testing all possible interactions. Therefore,
all possible two—way interactions were considered in the analysis. An ANOVA was run
with two-way interaction effects between the main structural design factors. No

signiﬁcant interaction effect was detected.

219

Table 6-17 Summary of p-values from ANOVA for determining the effect of design
factors on ﬂexible pavement performance—WNF Zone

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

Performance Measures
D ‘ Lon itudinal
F2216? Fatigue Rut IRI Transverse cricking
kin Depth cracking

“ac g AIRI IRlo WP NWP
HMA .
thickness 0.077 0.576 0.948 0.141 0.383 0.759 0.532
Base type 0.545 0.547 0.065 0.117 0.470 0.803 0.110
385° 0.703 0.476 0.144 0.559 0.806 0.937 0.265
th1ckness
Drainage 0.298 0.031’ 0.725 0.032 0.306 0.760 0.142
sue 0.000 0.000 0.008 0.000 0.276 0.037 0.000
locked)

R2=0.834 R2=0.561 R2=0.503 R2=0.680 R2=0.965 R2=0.662 R2=0.579

N=36 N=72 N=49 N=72 N=14 N=24 N=45

 

Also shows operational/practical signiﬁcance

Table 6-18 Summary of marginal means from AN OVA for determining the effect of
main design factors on pavement performance measures— WNF Zone

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

Average Performance
D . Fatigue Rut IRI Transverse Long1tud1nal
es1gn Factor . . crackmg
cracking depth cracking
(sq-m) (m) “R1 m (m) WP NWP
(m/kmL (In/kt“) (m) (m)
4,, 6.4 6.2 0.34 0 82 0.3 3.5 13.7
HMA (1.2) (1.78) (1.157) ' (1.412) (0.56Q (1.96)
Thickness 7,, 2.9 5.9 0.33 0 785 1.0 4.1 17.5
(0.43) (1.74) (1.165) ' (0.363) (0.73) (2.2)
5.5 5.9 0.40 0.6 3.7 19.7
DGAB (1.05) @744 (-0.989) 0'84 (0.872) M14) (2.32)
3.2 6.4 0.29 1.6 2.8 8.5
Base Type ”B (0.524) (1.815) (1.3) 0'81 (0.143) (0.356) (1.48)
ATB/ 4.4 5.8 0.32 0 764 0.2 5.2 21.8
DGAB (0.84) @312) (1.2) ' (1.93) Q97) (2.42)
8” 5.8 6.4 0.39 0 82 0.4 4.5 26.6
(1.104) (1.82) (1) ' (1.171) (0.812) (2.62)
Base 12,, 4.7 5.8 0.31 0 793 0.5 3.6 14.7
Thickness (0.89) (1.72) (1.23) ' (0.982) (0.603) (2.03)
16” 2.9 5.9 0.30 0 79 0.8 3.3 9.6
(0.42) (1.73) (1.26) ' (0.51) (0.523) (1.6)
N 3.3 6.6 0.57 0 833 0.9 3.4 11.1
Drainage (0.54) (1.85) (1.136) ' (0.385) (0.54) (1.75)
Y 5.8 5.2 0.53 0 772 0.3 4.2 21.3
(1.1) (1.6) (1 .2) ' (1.391) (0.752) (2.4)
MSE (1.3) (0.09) (0.146) (0.647) (1.37) (1.321)

 

Note: Values in parenthesis are the lognormal marginal mean values.

220

 

 

6.6.3.2 Effect of Site Factors on Pavement Performance

The third AN OVA was targeted at determining the signiﬁcance of subgrade type
and climatic zone. Trafﬁc, age and variability in target layer thicknesses Were considered 1
as covariates. The subgrade type and climatic zone were included as main factors in
addition to the structural design factors. For fatigue cracking, the analysis was run with
and without the Kansas (20) data, since the test sections in Kansas (20) have a large
amount of fatigue cracking and the project is known to have had construction problems
with a wet subbase and variable densities. The analysis for rutting was also done with and
without the Texas (48) data since rutting for these sections is believed to be due to the
asphalt mix [6].

The results ﬁ'om this analysis are summarized in Tables 6-19 and 6-20, and
generally show lower R2 values than those in Table 6—13. This may be partially due to the
effects of variations in environmental conditions within a given climatic zone and
variations in material properties within different pavement layers among sites. In
addition, construction and material variability is not accounted for in this analysis, since
the site factor is not blocked. The results seem to indicate that the effect of the climatic
zone is signiﬁcant for all performance measures and that the effect of subgrade type can
be signiﬁcant for most of them. However, caution must be exercised in interpreting these
results given the unbalanced nature of the design with respect to climatic zone: there are
only two projects in each of the dry zones, Dry-Freeze (DF) and Dry-No—Freeze (DNF),
as opposed to eight projects in the Wet-Freeze (WNF) and six projects in the Wet-No-

Freeze (WF) zones.

221

Finally, AN OVA was conducted by considering the main and interaction (two-way)
effects for all six experimental factors. The results of this analysis are summarized in
Tables 6-21 and 6-22. The conclusions are based on the main effects when interaction
between site factors is not signiﬁcant.

In case of signiﬁcant interaction between site factors, the interpretation of results are
based on the comparison of cell means, i.e., the mean performance of sections
corresponding to each subgrade type should be compared within each climatic zone.
The following discussion summarizes the effect of climatic zone and subgrade type on
the key performance measures:

0 Fatigue cracking: More fatigue cracking was observed on sections located in “wet”
climates. The interaction effect between subgrade soil type and climatic zone is
statistically signiﬁcant (see Table 6-21); therefore the conclusions are based on the
interaction effect. More cracking was observed in pavement sections built on ﬁne-
grained subgrade especially in WNF zone. Among the sections located in WNF
zone, the difference between the mean cracking of sections built on ﬁne-grained and
sections built on coarse- grained soil is also practically signiﬁcant.

0 Structural Rutting: Rutting was higher among sections located in “wet” climate and
was generally higher for pavement sections on ﬁne-grained subgrade. Both of these
effects statistically and practically signiﬁcant. There were high rut depths observed
in the Dry-No-Freeze (DNF) zone; however, it is believed that this is more related to

the asphalt mix as opposed to structural rutting.

222

Table 6-19 Summary of p-values from AN OVA for determining the effect of site factors
on avement performance measures (Main effects only)

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

Performance Measures
Site Factor Fatigue Rut IRI Tmsvmc Longitudinal crackirL
cracking Depth AIRI IRIO cracking WP NWP
Subgrade type 0.680 0.432 0.000‘ 0.067 0.020 0.015 0.013
Climatic Zone 0000’ 0000‘ 0.000' 0.001 0.000' 0.000 0.000
Trafﬁc Level 0.000 0.024 0.083 - 0.000 0.437 0.000
0.068 0.013 0.091 — 0.000 0.050 0.009
RT=O.288 R2=0.305 R2=0.401 R2=0.215 R2=0.674 R2=0.434 R2=0.534
N=124 N=159 N=163 N=212 N=67 N=95 N=134

 

 

Note: The model considered for this analysis has main effects for all six experiment factors. KS (20) was
not considered for analysis of all cracking measures, whereas, rut depth analysis was conducted

without TX (48). ‘ Also shows operational/practical signiﬁcance.

Table 6-20 Summary of marginal means from ANOVA for determining the effect of site
factors on pavement performance measures (Main effects only)

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

Performance Measures
, Lon itudinal
S1te Factor Fatilgue I)Ruth IRI T321351? cricking
crac mg ept AIRI IRIO WP NWP
1: 15.4 5.2 0.59 0 81 1.2 44 20.8
Subgrade (1.16) (1.6) (0.62) ' ﬂ38) (2.5) (2.2)
type c 18.5 5.4 0.44 0 76 4.9 13.2 42.0
(1.34) (1.65) (0.92) ' (1) (1.3) (2.9)
53.4 4.9 0.59 13.2 26.5 139
W (2.4) (1.56) (0.628) 0'86 (2) (2.0) (4.1)
24.0 5.6 0.336 1.5 16.1 25.4
Climatic WNF (1.6) (1.68) (1.186) 0797 (0.15) (1.5) (2.4)
Zone 26.5 4.0 0.58 3.2 5.92 56.5
DF (1.7) (1.34) (-0.641) 0784 (0.6) (0.5) (3.2)
4.5 7.1 0.596 0.4 177 3.12
DNF (0.08) (1.92) (0.613) 0695 (1.5) (3.9) (0.3)
MSE (3.156) (0.091) (0.19) (1.161) (2.556) (1.668)

 

 

 

223

Table 6-21 Summary of p-values from AN OVA for determining the effect of site factors
on pavement performance measures (With interaction effects)

 

Performance Measures

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

Site Factor Fatigue Rut 1R1 Transverse 1713?: a1
cracking Depth AIRI IRIO Cracking WP NWP
Subgrade type 0.626 0.886 0.018 0.653 0.247 0.480 0.191
Climatic Zone 0.049 0.007’ 0.000‘ 0.003 0.004’ 0.002 0.000’
Subgrade*Zone 0.000' 0.257 0.562 0000" 0.092 0.000" 0.496
Trafﬁc Level 0.031 0.028 0.150 - 0.197 0.655 0.000
Age 0.068 0.025 0.565 - 0.077 0.014 0.202
R2=0.575 R2=0.47 R2=0.525 R2=0.435 R2=0.931 R2=0.755 RT=0648
N=124 N=159 N=163 N=212 N=67 N=95 N=134

 

Note: The model considered for this analysis has main effects for all six experiment factors and all
#
possible two-way interactions between them. Also shows operational/practical signiﬁcance.

Table 6-22 Summary of marginal means from AN OVA for determining the effect of site
factors on pavement performance measures (Interaction effects onlj)

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

Performance Measures
Site Factor Fatigue Rut IRI Transverse Longitudinal cracking
cracking Depth AIRI [RIO cracking WP NWP
Subgrade type - - - - - - -
Climatic Zone - - - - - - -
22.4 60.8
F 3 WF F (1.7) ‘ ’ 0'95 ' (3.1) '
O
N C 67 - _ 0 76 _ 3.0 _
“:8 (2.8) ' (0.1)
a F 135 _ _ 0.78 _ 27.3 __
.o (3.5) (2.3)
a WNF
C 2.5 _ _ 0 83 _ 5.5 _
(0.5) ‘ (0.7)
MSE (2.816L (2.015)

 

 

 

 

 

 

 

 

 

 

 

Note: The cell means are only given when interaction is signiﬁcant. For main effects see Table 5-20 for
marginal means. ~

224

o Roughness: Both subgrade type and climatic zone are very signiﬁcant factors
affecting roughness growth (see Table 6-19). The pavements constructed on ﬁne-
grained soils have shown higher changes in roughness than those constructed on
coarse-grained soils. Also, pavements located in the WP zone have shown higher
change in roughness than those located in WNF zone. These effects were found to be
statistically and practically signiﬁcant (see Table 6-20). The effect of subgrade soil
appears to be mainly caused by the initial roughness being signiﬁcantly higher in
sites with ﬁne-grained subgrade. The initial IRI (IRIO) was found to be associated
with future roughness, especially among sections built on ﬁne-grained soils and
among sections located in “wet” climate.

0 Transverse cracking: The effects of subgrade type and climate on transverse cracking
are signiﬁcant. More cracking was observed in sections built on coarse-grained soil
compared to those built on ﬁne-grained soil. However, the magnitude of cracking at
this point in time is too low to conclude on the effect. More cracking occurred in
sections located in WF zone compared to those located in other zones, and this effect
was found to be statistically and practically signiﬁcant.

0 Longitudinal cracking: As the interaction effect between subgrade type and
climatic zone is signiﬁcant for longitudinal cracking-WP, the conclusions are based
on comparing cell means for sections built on each subgrade type within each
climatic zone. It was found that among pavements located in WF zone, those
constructed on ﬁne-grained soils have shown signiﬁcantly more cracking than those
constructed on coarse-grained soils. The effects of subgrade type and climate were

signiﬁcant in the case of longitudinal cracking-NWP. Signiﬁcantly more cracking

225

was observed in the sections built on coarse-grained soil compared to those built on
fme-grained soil. Also pavements located in “freeze” climate have shown

signiﬁcantly more cracking compared to those in “no-freeze” climate.

Given the unbalanced design of the experiment with respect to climatic zone, a
one-way AN OVA was performed to investigate the effects of subgrade type (fme-grained
versus coarse-grained soils) and climatic zones (wet versus dry, freeze versus no~freeze),
one at a time. The p-values and mean performances by site factors, from this analysis, are
summarized in Tables 6-23 and 6-24, respectively. To indicate the direction of effects for
site factors, the “+” and “—“signs are also reported along with the p-values. The “+”
indicates that, within a factor, the ﬁrst level is exhibiting more distress than the second

6‘ ‘6

level, while indicates otherwise. For example, in the case of the effect of subgrade on
fatigue cracking, the “+” indicates more cracking in pavements constructed on fme-

grained soils (ﬁrst level for subgrade) compared to those constructed on coarse-grained

soils (second level for subgrade).

The p-values indicate that subgrade type appears to be signiﬁcantly affecting
fatigue cracking, rut depth, roughness, transverse and longitudinal cracking-WP. The
pavements built on ﬁne-grained subgrade have shown higher distress than those
constructed on coarse-grained subgrade. The effect was found to be practically signiﬁcant
in the case fatigue cracking, rutting, change in roughness and transverse cracking. The

effects of site factors by performance measure are listed below:

226

Fatigue Cracking: Climate appears to be signiﬁcantly affecting fatigue
performance. Pavements located in “wet” or “freeze” climate have exhibited
signiﬁcantly higher amount of fatigue cracking than those located in “dry” or “no-
ﬁ'eeze” climate, respectively. This effect was found to be practically signiﬁcant.
Rut Depth: On average, rutting appears to be higher in wet climate. Also
pavements located in DNF zone were found to have signiﬁcantly more rutting
compared to those located in DF zone. However, it is believed that this is more
related to the asphalt mix as Opposed to structural rutting, as mentioned before at
the beginning of this chapter.

Roughness: Signiﬁcantly higher growth in roughness was observed for pavements
located in WF zone compared to those located in WNF zone. This effect was
found to be practically signiﬁcant.

Transverse Cracking: It was found that pavements located in WF zone have
exhibited signiﬁcantly higher transverse cracking than those located in WNF
zone. This effect was found to be practically signiﬁcant.

Longitudinal Cracking-WP: Signiﬁcantly more longitudinal cracking-WP was
observed in pavements located in WF zone compared to those located in WNF
zone. Also, signiﬁcantly more longitudinal cracking-WP was exhibited by the
pavements located in DNF zone compared to those located in DF. In DNF zone,
longitudinal cracking-WP and rutting is mainly contributed by sections in the
Arizona, AZ (4), site, where HMA-related issues are believed to be causing the

distresses.

227

0 Longitudinal Cracking-NWP: Signiﬁcantly more longitudinal cracking-NWP was
exhibited by the pavements located in WF zone compared to those located in
other zones. Also, more cracking was observed in pavements located in DF zone
compared to those built in DNF zone. These effects indicate that this distress

could be related to “freeze” environment.

It should be noted that the data from the four projects in the dry climatic zones
show negative trends in several performance measures. This may be in part due to the

lower number of projects in these zones.

6.6.3.3 Effect of Design Factors on Pavement Performance (univariate) based on
standard deviate

As explained before, the experiment design and the performance of the test sections have
rendered the SPS-1 experiment “unbalanced”. Fourteen out of eighteen sites, in the
experiment are located in “wet” climate, of which eight are in the WF zone. In addition,
all 24 unique designs were not built in every soil-climate combination. Furthermore, non-
occurrence of distresses in a considerable number of sections contributed to the
unbalance. This could be a reason for insigniﬁcance of interaction effects between the
design and site factors from multivariate analyses presented above. In light of the above
concerns, a simpliﬁed analysis considering one design factor at a time (univariate) was
performed using one-way ANOVA (as in the case of analysis of the effects for site

factors).

228

Table 6—23 Summary of p-values from one-way ANOVA for determining the effect of '

site factors on pavement performance measures

 

Performance Measures

 

 

 

 

 

 

 

 

 

 

 

 

 

. . Longitudinal
Slte Factor F atrgue Rut Transverse crackin
cracking depth AIRI IRIO cracking WP NWP
Subgrade Type
Fine vs. Coarse 0.03 (49' 0.002(+)' 0.00 (+)' 0.01 1 (+)‘ 0.016 (+)‘ 0.001(+) 0.26 ()
Climatic Zone
Wet vs. Dry 0.021 (+)‘ 0.087 (+) 0.596 (-) 0.000 (+)‘ 0.005 (+)‘ 0.919 (+) 0.040 (+)
r vs. NF 0.011 (+)’ 0.001 () 0.000 (+)‘ 0.010 (+)’ 0.060 (+)‘ 0.038 () 0.000 (+)
wr vs. WNF 0.063 (+)' 0.893 () 0.000 (+)' 0.030 (+)' 0.000 (+)‘ 0.096 (+) 0.000 (+)
DP vs. DNF 0.054 (+)' 0.000 (-) 0.281 () 0.231 (+) 0.055 () 0.000 () 0.001 (+)

 

i
Also shows operational/practical signiﬁcance

Table 6-24 Summary of marginal means from one-way ANOVA for determining the
effect of site factors on pavement performance measures

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

Performance Measures

. , Longitudinal

Slte Factor Fatigue Rut IRI Transverse crackinL
cracking depth AIRI [RIO cracking WP NWP
Subgrade Type Fine 54.2 5.8 0.613 0.86 18.97 64.5 163.0
Coarse 24.2 4.8 0.451 0.79 6.70 17.4 113.0
Wet 41.7 5.4 0.524 0.85 15.10 39.7 150.5

Dry 17.4 4.7 0.552 0.73 04.83 38.1 74.0
Freeze 45.6 4.8 0.616 0.85 13.57 24.5 189.7

Climatic Zone No Freeze 18.4 5.8 0.425 0.78 06.20 57.7 27.1
WF 53.7 5.2 0.635 0.88 24.3 31.3 244.7

WNF 24.5 5.6 0.360 0.81 2.80 14.8 29.4

DF 26.2 4.0 0.480 0.76 2.25 4.10 83.1

DNF 07.8 6.8 0.572 0.70 6.00 108.8 14.7

 

 

 

229

 

The performance of test sections was not found to be consistent across sites indicating the
inﬂuence of site conditions (see Chapter 3). The site conditions that could have
contributed to this variation in performance are trafﬁc, age, construction quality,
measurement variability, and material properties, apart from the experimental site factors
(i.e. subgrade and environment). In order to separate the “true” effects of the
experimental factors, this “noise” had to be nulliﬁed.

The standard deviate for each performance measure was calculated, within each
site, for all the sections using equation 5-3. This measure indicates the relative
performance of a design compared to the other designs. As this measure was calculated
for each section, considering one site at a time, it indicates the relative standing of the
section compared to other sections. It thus helps nullify the variation in performance (due
to site conditions) among sites, as the sections are weighed with respect to companion

sections in each site.

Std .Deviate = (M) (5-3)
0'

The above approach of using the standard deviate is similar to blocking of the site
factors performed in the multivariate analysis. One-way analysis of variance (univariate)
was performed on the standard deviates of the sections to study the effects of each design
factor by taking one design factor at a time. The analyses were performed on data from
all sections and also on subsets of data stratiﬁed by different subgrade types, climates and
combinations of these. This helps identify the effects of design factors under different site
conditions.

In the SPS-1 experiment, HMA thickness and drainage have two levels (i.e. 4” vs.

7” and drainage vs. no drainage). But for base thickness and base type, three levels (8”

230

vs. 12” vs. 16” and DGAB vs. ATB vs. ATB/DGAB) are present. Moreover, 16” base
thickness was provided only for drained sections and ATB/DGAB was built only for un-
drained sections making the design unbalanced. Therefore, for studying the effects of
base thickness and base type, analyses were done separately among drained sections and
among un-drained sections.

To see the “pure” effect of each design factor, comparisons of standard deviates
were also made between the levels of each design factor while controlling the other
factors, as in the case of level-B analyses (site-level). The results from this analysis are in _
Appendix A5 of reference [7].

The effects of design factors, based on the above-mentioned analyses, on each

performance measure are discussed next.

Fatigue Cracking

The effects of the design and site factors, in terms of standard deviate, are shown
in Figure 6—22. In addition, the summary of p-values corresponding to the analyses
performed to study the effects of design factors on fatigue cracking is shown in Table
6-25. The mean area (m2) of fatigue cracking (PI) corresponding to each comparison
presented in Table 6-25 is shown in Table 6-26. Though the univariate analyses were
performed on standard deviates, the mean cracking was used to identify operationally 1
signiﬁcant effects. The effects of design factors on fatigue cracking, based on this
analysis, are presented below:
HMA thickness: The effect of HMA surface thickness is statistically and operationally

signiﬁcant, especially among sections located in WNF zone. Sections built with “thin”

231

[4-inch (102 mm)] HMA surface have exhibited higher fatigue cracking than those built
with “thick” [7-inch (178 mm)] HMA surface. On average, among sections built on ﬁne-
grained soils, higher fatigue cracking was observed on “thin” [4-inch (102 mm)] sections
compared to “thick” [7-inch (178 mm)] sections. However, this effect is not signiﬁcant.
Similar trend was found among sections built on coarse-grained soils and the effect is
statistically and operationally signiﬁcant.

Base thickness: The effect of base thickness is marginal among sections built on ﬁne-
grained subgrade soil, in that sections with thick 16-inch (406 mm) permeable base have
exhibited lesser cracking than those with 12-inch (305 mm) or 8-inch (203 mm) base
thickness. Also among sections located in WF zone the effect of base thickness on
alligator cracking is statistically and operationally signiﬁcant. Sections built with 16-inch
(406 mm) permeable base have exhibited lesser cracking than those with l2-inch (305
mm) or 8-inch (203 mm) base.

Base Type: The effect of base type (unbound versus treated base) is statistically and '
operationally signiﬁcant, with ATB giving the “best” performance and DGAB showing
the “worst” performance. This effect is more prominent among sections built on ﬁne-
grained soils compared to sections built on coarse-grained soils. Also, this effect is more
noticeable among sections located in WF zone.

Drainage: The effect of drainage is signiﬁcant (statistically and operationally) among
those in WF zone and built on ﬁne-grained subgrade soils. Sections with drainage have
lesser cracking than those without drainage. This effect is more prominent for the

sections constructed with DGAB than those with ATB. This shows that drainage is more

232

effective if provided with DGAB than when with ATB. The interaction effects among the
experimental factors, on fatigue cracking, are reported below:

In general, “thin” sections with DGAB on ﬁne-subgrade soils have exhibited the
most alligator cracking while “thick” sections with ATB on coarse—grained subgrade soils
have exhibited the least alligator cracking. Among un-drained pavements, on average, an
increase in HMA surface thickness from 4-inch (102 mm) to 7-inch (178 mm) has a
slightly higher effect on fatigue cracking for pavements with DGAB than for pavements
with ATB.

Among sections located in the WF zone, those with DGAB have shown the
highest amount of cracking while those with ATB have the least. In addition, among
pavements located in WF zone, those with l6-inch (406 mm) drained base have less
fatigue cracking than others. These effects were found to’be statistically and practically
signiﬁcant. Among pavements with DGAB and built on ﬁne-grained soils, those with
drained base have lesser fatigue cracking than others. Also, among sections with drainage
and built on ﬁne-grained soils, those with 16-inch base have lesser cracking. These

effects were'found to be statistically and practically signiﬁcant.

233

Mean std. deviate

Nban std. deviate

Mean std. deviate

 

 

 

 

 

 

 

 

 

 

 

1.2 E
0.7 E
0.2 E. _ I I _
-0.3 E
-0.8
t
-1.3 ”
2 2 2 :: e: 3 < 3 g < 2
So a} _ _. E
. E <2 <
DemgnFactors :14
(a) Overall
1.2 : .
DFineICoarse
0.7 :
t
0.2 :E. I. IL
n 1J1 I' t 1
033 U U U- D H D
-0.8 E
t
-1.3
it 11 ’2 A Q 9’: 5 E E D
.2 2 so is: :7; a < < E
. < < <
DesrgnFactors 9*
(b) By subgrade type
1.2 :
; DWFIWNF
0.7 :
0.2 El H ;l _
: 1 I T IL] I
~O.3 :
-0.8
-1.3
a.» a c a a o a e E o
.2 é a; 5: r: 5 < < 9
<
< E .

Design Factors
(o) By zone type

Figure 6-22 Effect of design factors on fatigue cracking

234

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

.ooconcaou do 3?... conmE no $3 a ESE—Em b32538 26%. 2.8 eocezm ”202
cm VN 0m 3.. VN VN oo ow we on ww— Z
oomd OVNd O26 0:.0 omNd oomd ooﬁo oomd owwd 0010 oMNEED-OZ .m> ems—:30 mCt<
ovwd omvd oomd ONNd Ohmd ownd ciao .omeEDéZ. .m> umeED m<OD emu—:50
on; on; 82 83 $3 and See :8 6858er .9 ceases :85
52.5: .9 E<a=m<oo woman
omwd ObNd ooNd Nmmd mw_.o .m> m<0n<mir<
.2, m5 .2 m<oo =<

owed chNd 3:6 m_m.o 2:6 52.0 m._.< .m> m<OD Q cab ommm

. . . . . . 385:
E o Em C as. o a; o o: o SN o ,9 a: as 38 oz

. . . . m<oQE<
o2 o o: o com o can o as at. .9 98c :85
hand homd mNnd mm—d hhad oowd 52.0 N—nd gmd aoﬁ .m> :N_ .m> an O

m
E2 88 ES :26 we; :2 82. 83 55 as .2 ..w oz mwmume
omvd omwd memo mwd o—vd omwd :Nd :0— .m> :N— .m> :w =Eo>0
Ago—.0 oomd Goad 2N6 oowd “.26 :h .m> :v wave—£5 <2:
0 r.— U _ m U m U m
.mZD LO “~23 m3 uni—~00 DEE
LZD mo "—23 m3 =Eo>o sowing—coo cocoa..— amino
ucou use 3233 >m o=o~ Dannie am 03533 Am

 

 

 

 

 

 

 

wagons oswwmm 1.333% c.8283 mo 203388 How 853?.” mo EBB—5m mmie 2an

235

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

236

 

 

 

 

 

w.m o.o ”2 o._ v.3 6v o6 v; ”.2 _.m mé ad a.» N6 0
mm ..o .0: Yo m6 Qw a6 5v 0: Wm us 56 ed we DZ mH<
Em od oém M: mow me m6 wN oém :.— c.: :2 QE 5.: Q 0

. . . . . . . . . . . . . . m<OQ oemEEQ
NN. mo 92 v_ m: mmm m: mm a2 ooM com 02 new mom DZ
N6 o6 vdm v._ Won m6. we ..m 13 _.m_ 3: od— bﬂ— m.m_ Q 80>
we Nd Q: wd men v.3 m.w m.m Q: ..2 0: Wm v.2 0.: DZ = O
wN c6 ”.3 o._ ﬁlm 5v o6 v; no. mo men Qm a.” N5 mh<
56 ed oém mg mow me ad ad oém :L 0.: _.w_ QE 5.: m<OD Q

:4 od 2: We me 0m N.» 9m 2: m.: Wm Nam v.2 w.m m<09mrr<
mm ..o 0: Yo ms ma 0.0 he. 0: Wm 56 he 0.0 v.0 m._t< 02 8b ommm
m6 m6 o.m_ v; 03. mwm 0: w.m ad. 93 odm o.m_ mém odm m<OQ

..v o.o 2: m6 me Wm Wm 0N md. 0: Na Nam v.2 ad m<OQ\m:.<
Ym od 5.3 m6 wh— ee mum 5N 5.3 we ma N6 55 we mh< =Eo>O
me No odm h; wdm NHN 56 NM Qom mNm Of ..2 YMN M3: m<OQ
m.» o.o m._N v; _.mm ad QN m6 m._m 56 Pm Wm M: M; :2
Yo o.o 0.0m 0.. m._m w.: W” No Qom 0.5 2: 2: v. "N Q: ..m_ D
w.m _.o oﬁm N; S: v.3 new 0.. aﬁm ﬁx 9: m6. wd v.2 em
O: «.0 ad Q. :2 0.9 m.” ow Qm mm as: v.0 2: mm :9

. . . . . . . . . . . . . DZ 3.65.25 095
mm oo 92 co :m .a mm : om: ~NN \tw wh mmm cm. ..w
m.» 0.0 m._m v; _.Nm ad o.m Nd m._m 56 gm m6 as M; :3
06 md WE N; w._m Wo— vw 9m m4; Wm. v.9 dd 3: Q: em— :Eo>O
aim od mgm wd ..ov v.2 a6 v; m._m 02 ”I ma: N: v.2 ..w
m6 o.o 92 9o mdm Wm. v.5 0m 92 v.» N: 5w 2: ed as

. . . . . . . . . . . . . . $05.25 <2:
no No new m: one mm. mm vm mvm mm. 2: ”2 N2 03 ..v

0 m U m U a. U r.— mZQ no "—23 m3 3.800 0:3 .

h~20 mm "—23 m3 :Eo>O cemtaanU .coaomm :wrﬁﬂ
0:3 "ESE—o mm onﬂwnnm tam
econ can ouﬂwnam mm

 

 

 

 

 

 

mcﬁoﬁo ozwuﬂ com 5 MO 338 .«o baﬁﬁsm £6 3an

 

Structural Rutting

The effects of the design and site factors, in terms of standard deviate, are shown
in Figure 6-23. The summary of p-values corresponding to the analyses performed to
study the effects of design factors on structural rutting is presented in Table 6-27. The
mean rut depth (PI), in mm, corresponding to each comparison presented in Table 6-27
are shown in Table 6-28. The effects of design factors on rutting, based on this analysis,
are presented below:
HMA thickness: Among sections built on coarse-grained soils, the sections built with
“thin” [4—inch (102 mm)] I-[MA surface have exhibited higher rut depths than those built
with “thick” [7-inch (178 mm)] HMA surface. This effect is statistically signiﬁcant but
not Operationally signiﬁcant, at this point in time. Thus increasing HMA thickness from
4” to 7” may be more effective in retarding rutting in the case of sections with coarse-
grained soils than in the case of sections with fme-grained soils. On average, sections
built on ﬁne-grained soils have slightly higher rutting than those with coarse-grained
soils.
Base thickness: The effect of base thickness is signiﬁcant (statistical and operational)
among sections located in WNF zone where higher rutting was observed for the sections
built with 8-inch (203 mm) thick base than for those built on 16-inch (406 mm) thick
base. In addition, this effect seems to be more apparent for the sections built on coarse-
grained subgrade soils.
Base Type: In general, the effect of base type (unbound versus treated base) is not

statistically signiﬁcant. However on average, sections built on ATB have shown the

237

better performance than those sections built on DGAB. This effect (DGAB vs. ATB) is
more prominent among sections located in WF zone and built on fme-grained soils.
Drainage: In general, the effect of drainage is statistically signiﬁcant with un-drained
sections showing higher rutting than those with drainage. However, this effect is not
operationally signiﬁcant. This effect is signiﬁcant (statistical and operational) among
sections located in WNF zone and builton fme-grained soils. Also the effect is
signiﬁcant (statistical and operational) among sections in WF zone and built on coarse-
grained soils. The results suggest that drainage may be more effective in inhibiting rutting
for pavements on ﬁne-grained soils, when located in WNF zone.

The interaction effects among the experimental factors, on structural rutting, are
reported below:

A marginal effect of drainage was observed on pavements built with ATB and on
ﬁne-grained soils. Also, among drained pavements located in WF zone, those with
DGAB have shown higher rutting than those with ATB. Furthermore, among sections
located in WF zone and built with ATB, those with drainage have shown signiﬁcantly
less amount of rutting than those without drainage. Both of the above effects were found
to be statistically signiﬁcant and are of Operational signiﬁcance.

Among un-drained sections located in WNF zone, those with 12-inch (305 mm)
base thickness have less amount of rutting than those with 8-inch (203 mm) base
thickness. This effect was found to be statistically signiﬁcance and is practically
meaningful. For sections built on DGAB and located in WNF zone, those with drainage
have shown slightly lesser rutting than those without drainage. The effect was not found

to be statistically signiﬁcant.

238

Wan std. deviate

Mean std. deviate

Mean std. deviate

1.0

0.8

0.5

0.3
0.0
-03 :

-0.5
-0.8
-1 .0

1.0
0.8
0.5
0.3
0.0
-0.3
-O.5
-0.8
-l .0

1.0 :
0.8 E
0.6 ;
0.4 E
0.2 i
0.0 E
-o.2 :
-0.4 f
-0.6 i

-0.8

 

 

 

 

 

 

4"

7H

8" (ND)
12" (ND)
3" (D)
12"(D)
16"(D)

“iéii

Design Factors

(a) Overall

 

TIWII‘T'TT‘

WWI—TY

El Fine I Coarse

7

 

 

7"

3" (ND)
12" (ND)
3" (D)
12"(D)
16"(D)

Egg

Design Factors

PATB/DGAB
ATB/PATB

(b) By subgrade type

 

 

 

 

 

4"

a " " a 5 5 E E.“
3? are §<§§g
°° :3
<
DesignFactors < E

(0) By zone type
Figure 6-23 Effect of design factors on structural rutting

239

.oocovuzoomo _o>o_ 8:3: 8 goo an “ESE—Ea Diuumzﬁm >35 m=oo eucacm ”8oz

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

VN mm on On N— VN @n 00 V” g ~0— Z
A: Md owmd omNd OhNd ON _ d OWNd on _ d owaEEQéZ .m> omeED m—P<
omnd A: _d onmd ddmd O—hd 02d ovod owwd 005d o—wd awaEEQéZ .m> own—EEO m<OQ emu—EEO
dh—d 005d owmd O”; .0 A: —d uMaEEQ-OZ .m> omaEEQ :Eo>O
mF<m\m:.< .m> mF<n<m<8 mummm
whmd movd mde E Nd mend whNd d: d .m> m<8mh<
.9 m: .9 950 :<

mend ow — d Odvd hwwd :.vd Nmod A: _.o Nmmd mk< .m> m<OD D 090 umnm

. . . . . . . . . . m<oQE<
MS c an o com o § o N8 o SN 0 :o o E. o a; o 3m 0 .2 E< .: $8 oz

. . . . . . . . m<oQE<
O: o on.— d Obm o CNN 0 ONW o EN 0 mNN d om: o .m> m._.< .m> m<OD =Eu>O
wN_ d wnmd wavd Vwbd Nﬁhd mood mde homd mmod wmvd 2.6.0 :0— .m> :N_ .m> :M D

82 on _ .o ‘ $3 :3 $2 9.3 «2 .o ..N. .e, ..w oz mmwﬁms
oomd oomd coed omed ommd owmd ovmd Omwd OhNd am: .m> :N_ .m> :w :Eo>0
oomd Good Good omvd —omd dwwd SNd wwvd :b .m> :v 305—2:. <2:
0%... o _ m o m o m
mZQ LO m2? E5 3.500 0::
mZQ ...5 m2? m3 =Eo>O sewing—:00 86mm cwioo
BEN Ea 2.65:3 am neon "Ea—£3 ‘3 09233 mm

 

 

 

 

 

 

 

maﬁa 3.588% lmoumgow wawgdm mo 383368 How 32gb mo SEEsm Rb 055.

240

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

3 3. 3 3 3 3 3 3 3 3 3 3 3 o
3 3 3 3 3 3 3 3 3 3 3 3 3 oz m3

3 3 3 3 3 3 3 3 3 3 3 3 3 o

3 3 3 3 3 3 3 3 3 3 3 3 3 oz 30o 3:35
3 3 3 3 3 3 3 3 3 3 3 3 3 o

3 3 3 3 3 3 3 3 3 3 3 3 3 oz :35

3 3 3 3 3 3 3 3 3 3 3 3 3 m3

3 3 3 3 3 3 3 3 3 3 3 3 3 30o o

3 3 3 3 3 3 3 3 3 3 3 3 3 309m:

3 3 3 3 3 3 3 3 3 3 3 3 3 m3 oz

. . . . . . . . . . . . . 0903mm
3 3 3 3 3 3 3 3 3 3 3 3 3 30o

3 3 3 3 3 3 3 3 3 3 3 3 3 309m:

3 3 3 3 3 3 3 3 3 3 3 3 3 m? :35

3 3 3 3. 3 3 3 3 3 3 3 3 3 30o

3 3 3 3 3 3 3 3 3 3 3 3 3 ..2

3 3 3 3 3 3 3 3 3 3 3 3 3 ..o o

3 3 3 3 3 3 3 3 3 3 3 3 3 ..w

3 3 3 3 3 3 3 3 3 3 3 3 3 ...2 oz 322233
3 3 3 3 3 3 3 3 3 3 3 3 3 ..w .
3 3 3 3 3 3 3 3 3 3 3 3 3 ..2

3 3 3 3 3 3 3 3 3 3 3 3 3 ..2 :35

3 3 3 3 3 3 3 3 3 3 3 3 3 ..w

3 3 3 3 3 3 3 3 3 3 3 3 3 .s 305.0222:
3 3 3 3 3 3 3 3 3 3 3 3 3 ..v .
o o o o m o o uzo mo oz)» E5 838 2.:
mZQ mm “—23 m3 :So>O cemtmoEoU 958m“. 3500

econ 038:0 Am 3833 Am

 

 

 

 

 

33 new 3833 >m

 

 

 

 

 

 

 

wag: 358:? 8% 5 mo 338 no 9383 wmb 33%

241

Roughness
The effects of the design and site factors, in terms of standard deviate, are shown

in Figure 6-24. The summary of p-values corresponding to the analyses performed to
study the effects of design factors on roughness is shown in Table 6-28. The mean PI
corresponding to each comparison presented in Table 6-28 is shown in Table 6-29. The
effects of design factors on roughness, based on this analysis, are presented below:
HMA thickness: In general, the effect of HMA surface thickness is statistically
signiﬁcant. Sections built with “thin” [4-inch (102 mm)] HMA surface have exhibited
higher change in roughness than those built with “thick” [7-inch (178 mm)] HMA
surface. This effect is more prominent among sections built on ﬁne-grained soils.

Base thickness: On the whole, the effect of base thickness is statistically and
operationally signiﬁcant. Sections with “thick” [l6-inch (406 mm)] permeable base have
exhibited the least change in roughness whereas sections with “thin” [8-inch (203 mm)]
base thickness have shown the highest change in roughness. This effect is more
signiﬁcant among sections located in “wet” climate than among sections located in “dry”
climate.
Base Type: In general, the effect of base type is statistically and operationally signiﬁcant.
Sections built with DGAB have exhibited higher change in roughness than those built
with ATB. This effect is more prominent among sections built on ﬁne-grained soils.
Drainage: By and large, the effect of drainage is only statistically signiﬁcant i.e. it is not
of practical signiﬁcance at this pointin time. Sections without drainage have exhibited
higher change in roughness than those built with drainage. This effect is signiﬁcant

(statistical and operational) among sections built on ﬁne-grained soils and located in WF

242

zone. This effect is more prominent for sections with DGAB. This suggests that drainage
is more effective for pavements with DGAB on ﬁne-grained soils, especially when in WF
zone. The interaction effects among the experimental factors are reported below:

Also for un-drained pavements built on ﬁne-grained soils, the effect of base type
is signiﬁcant, in that pavements with ATB have signiﬁcantly lower AIRI. Furthermore,
the effect of drainage for sections with DGAB and built on ﬁne-grained soils, is
signiﬁcant. The above effects were found to be statistically signiﬁcant and are of
practical signiﬁcance. For un-drained pavements built on coarse-grained soils, an
increase in base thickness from 8-inch (203 mm) to 12-inch (305 mm) has a marginally
signiﬁcant effect, in that sections with thicker base have lower AIRI. However, this effect
is not of practical signiﬁcance at this point in time. It should be noted that, in general,

pavements built on ﬁne-grained soils have shown higher AIRI than those built on coarse-

grained soils, especially among sections in WF zone. Also, AIRI among sections located
in WF zone is higher than those in WNF zone.
Transverse Cracking

The effects of the design and site factors, in terms of standard deviate, are shown
in Figure 6-25. The summary of p-values corresponding to the analyses performed to
study the effects of design factors on transverse cracking is presented in Table Q-31. The
mean PI corresponding to each comparison presented in Table 6-31 is shown in Table
6-32. The effects of design factors are presented below:
HMA thickness: The effect of HMA surface thickness is not signiﬁcant. However, on an
average, sections with “thin” HMA surface have slightly higher cracking than sections

with “thick” HMA layer.

243

Mean std. deviate Ivan std. deviate

Wan std. deviate

 

1.0 E

 

 
   

 

 

 

 

0.8
0.6
0.4
0.2 f
0.0 g
-O.2
-0.4
-0.6 :
-O.8
-1.0
s s c c e a c E E a e
.2 2 so :‘l 50 3 < <
<
DesignFactors < E
(a) Overall
1.0 -
0.8 : CJFine ICoarse
0.6 E

 

   

 

 

 

-0.2 -
-0.4 5
-O.6 E
-0.8 E
"'0 ' ' e a e e
.. e c ” é: a c e
°° 9'.
<
DesignFactors < é
(b) By subgrade type
DWFIWNF

 

 

 

4"
7”

8" (ND)
12" (ND)
8" (D)
12"(D)
16"(D)

33;

Design Factors

(0) By zone type
Figure 6-24 Effect of design factors on change in IRI

PATB/[X3148
ATB/PATB

244

 

.oooovcooo .«o .26. .032 8 $9

in E8533 £8533 >35 m=uo 332m ”802

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

  

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

8 8 on 3 8 8 8 8m 2: Na 88 z
83 83 83 83 23 83 83 83 83 83 83 owgeoazaiwnsso o3
85 83 83 . 83 83 9.3 83 umu=_eosz.m2m2_eo o<oo owaaao
83 33 83 83 $3 83 03523232323 :35
, 5.253 .5, oh<n<o<oo 83mm
83 83 33 , 3.3 3.0 83 :3 .2303:
.. .:E<.m>o<oo =<
S3 83 23 E 83 £3 33 on; 33 23 E<.:m<oo o 3328
. . . 9:. . . . . . m<8mh<
23 $3 33. , .83 38° ”:5 83 88° .:E<§m<oo oz
. . . . . . . . m<8ﬂk<
SS 8; $3 2; 23 8; 83 83 .52.:38 __e..;o
83 $3 53 $3 33 £3 83 33 M33 M33 33 ..2.m>..~_.m>..w o
m
83 88.0 35 9.3 $3 83 :23?” oz mwmwms
9.3 83 83 83 83 83 ..E.m>..§.m>..m :35
83 83 83 83 83 83 3.3 83 Rare 385.0222:
o _ o o _ o o o o o
ozo mo .23 m3 038 of
"~29 “a "—23 m3 :Eo>O :83an0 Stan amino
2.3 as 0333?». am econ ogEzo hm ouﬁwnam zm

 

 

 

 

 

 

9: E 0930 [833% whvoﬁm mo 381388 com moigd mo bmﬁﬁom 36 053.

245

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

E d _md wdd odd Vdd Ed mmd mmd wdd mdd mmd Ed. 3d Ed . Q
Ed dmd Ndd odd odd Ed Sad de mdd odd dmd Ed Rd mmd DZ E4.
:d mmd Ed wdd mmd wmd dvd mmd Ed Ed mmd Ed mmd mmd Q m<OQ 36:65
E d mmd hmd bdd dmd Ed Nd; 3d hmd Ed odd Ed Ed 3d OZ
Ed de Ed Edd Ed mmd mmd mmd Ed Ed dmd Ed de dmd D :Eo>O
Ed 2d Ed ddd Ed Ed vcd mmd Ed d_d Ed Ed Ed dmd DZ
Ed :...d wdd ddd vod Ed mmd mmd mdd mdd mmd :.d .Nd E d mh<
:.d mmd Ed wdd mad and dvd de Ed Ed mmd Ed and mmd m<OQ 0
End dmd hdd Ed Vdd Ed .md dmd Sd Ed dmd Ed wmd dmd m<09mh<
Ed dmd Ndd ddd ddd Ed 3d de mdd odd dmd E d Sud mmd mb< DZ 85 ommm
Ed mmd Ed Edd dmd Ed Nd; End nmd Ed 8d Ed Ed :.d m<OQ
_md dmd ndd Ed Vdd Ed Ed dmd hdd d_d dmd Ed and end m<OD\m._.<
Ed wmd mdd odd mdd Ed End nmd mdd odd mmd :d Rd Ed 9:. :Eo>O
Ed 3d Ed hdd vmd vmd mod had Ed Ed tad Ed dvd .md m<OD
Ed 3d wdd mdd wdd Ed Ed mmd mdd odd mmd :d mmd Ed ..E
bdd dmd Ed odd vod mmd 3d Ed Ed odd and Ed dmd _md ..E D
Ed 3d :.d odd mmd dmd dmd de :d Ed dmd n _ .d dmd mmd ..w
Ed 9d Ed hdd mdd Ed de dmd Ed wdd End Ed wmd vmd ..E
. . . . . . . . . . . . . . DZ 3056:: 25
So So ddd So So vmd Ed emd ddd Ed dmd Ed dmd mmd ..w
vmd 3d wdd mdd mdd Ed wmd mmd wdd odd Ed 2d Rd E d ..E
Ed End Ed odd mdd Ed 3d mmd E d mdd mmd Ed cmd de ..E =Eo>o
E d mvd Ed 2d vmd dmd dmd dmd d_.d Ed Ed Ed Ed dmd ..w
Ed dmd Vdd wdd mdd vmd wvd vmd vod bdd hmd Ed End mmd ..E.
. . . . . . . . . . . . . . Q8535 <2:
2d dmd Ed wdd dmd Ed mmd _md Ed Ed and Ed Ed Rd ..E.
U m U m U m U m ”E20 "5 m2? E5 8:80 uni
LZQ "5 m2? “5? 0:3 26.53 am odﬁwdsm Am =Eo>O cemtdeoU £903 amino
gen ES 3833. km .

 

 

 

 

 

 

HE E omega 8.“ E do .888 do waE=m one 2de

246

 

Base thickness: The effect of base thickness is marginally signiﬁcant among sections
located in WP zone, especially among sections built on ﬁne-grained soils. Sections built
with 16” base have shown the least cracking while sections with 8” or 12” base have
shown the highest cracking.

Base Type: On the whole, the effect of base type is statistically signiﬁcant. Sections with
ATB have exhibited the least cracking while sections with DGAB have shown the
highest cracking. However, this effect is not operationally signiﬁcant at this point in time.
Drainage: In general, sections with un-drained sections showing higher cracking than
those with drainage. In addition, this effect is signiﬁcant (statistically and operationally)
among sections built on ﬁne-grained subgrade and located in WF zone.

On the whole, at this point in time, sections in WF zone have shown higher
cracking than those located in WNF zone indicating that transverse cracking is associated
with low temperatures. Also, among drained pavements built on coarse-grained soils, I
those with ATB. performed better than those with DGAB. However, among pavements
with DGAB and built on ﬁne-grained soils, those with drainage have shown signiﬁcantly
less transverse cracking than those without drainage. These effects were statistically
signiﬁcant and are of practical importance.

Longitudinal Cracking- WP

The effects of the design and site factors, in terms of standard deviate, are shown in
Figure 6-26. The summary of p-values corresponding to the analyses performed to study
the effects of design factors on longitudinal cracking-WP is presented in Table 6-33. The
mean PI corresponding to each comparison presented in Table 6-33 is shown in Table

6-34.

247

Mean std. deviate Mean std. deviate

Nban std. deviate

0.8

0.6 ‘

0.4
0.2
0.0
-O.2
-0.4
-0.6

-0.8 ‘

0.8
0.6
0.4
0.2
0.0
-O.2
-0.4
-0.6

0.4
0.3
0.2
0.1
0.0
-O.l
-O.2
-0.3
-0.4
-O.5

 

 

 

 

5. e :3 a a a ,. m E e a
6 E. 5 if 3 g < E g <: 3
30 E! '_ "' E E
. < < <
DesrgnFactors a.
(a)Overall

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

E
E
E
5. s " " 6 " " f2 E3 Q
2 3 s: 2 5: § < g g g Q
60 a. "' "'
<
DesignFactors < E
(b) By subgrade type
- CIWFIWNF
L
E
i: R A A 5 5 5 B O D
3 3 g; g: g g E < Z
30 a. — ~ E
<
DesignFactors < E.
(c) By zone type

Figure 6-25 Effect of design factors on transverse cracking

248

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

.uocoucgo .«o o>2 .553: .0 $8 an 285:»? £8338.“ 30% £8 392m ”202
VN VN N~ N— VN VN wv VN Nb xv o2 Z
OVNd o — Md owmd omwd o ~ Nd owmd om _ .o o” —.0 03580.02 .m> uMaEEQ mrr<
Omwd oo—d Obmd 02 .o o~o.o 05—6 oohd ommd O~N.o omeEn—éz .m> own—EEG m<OD quEED
ow _ .o O»; .o ommd Ommd 0:6 Ommd o_o.o owaEEQéZ .m> ummEEQ =Eo>O
mF<n<mF< .m> QF<EQ<OD mommm
vad ade nw—d 50— .o 5050 mde Nde .m> m<OQ\m._.<
.2 m: .2 m<oo :<

23 a; om. .o :3 $8 83 03° .5 .2 $8 o 25 3am

. . . . . . . . . . 3853
V30 wvwo nmmo goo Obmo vmwo Ohho 200 n—No Nh— o .m> m.~<.m> m<oo DZ

. . . . . . . _ 309m:
Em o 2m o om. o 08 o 8” o mm o , ea c .2 E< .2 38 .526
wcmd wNwd ~mm.o mvmd wwod mend Naad $06 waod mvwd V—md :©_ .m> :N— .m> :M Q

, mmD: U
m: .0 me .o 2:. . $3 093 $3 83 32 as .9 3 oz ammz.
ovmd OONd GENO cwwd movd Nde mmmd oﬁvd 936 :0— .m> sm— .m> 2w =n5>O
ovod 9:6 966 Omwd o— Md 2 Md 290 ommd 0:5 :5 .m> ..v 305—25 <2:
0 _ m o _ a u m o a
LZQ no "~23 mg 3.500 05n—
mZD "5 m2? m3 :Eo>o comtun—Eoo .98"— amino
28s “En ucﬁwnsm am 2.8 0:25? am 3333 am

 

 

 

 

 

 

 

wcioﬁo 35355. .ImonFoc wavaﬁm mo 385388 com mo:_m>-m mo baﬁﬁzm :6 39¢.

249

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

w; dd vd dd ad _.m ad— ad vd vd d.a ad _.m M: D
m._ dd md wd dd v6 Nam ad md ad Qw— v._ ..a 9m 02 mra<
02 dd ma dd ad Om. d.a Na ON nd _.: dd Nd mé D
. . . . . . . . . . . . . . m<oo amazes
com _d M: do a_ ad mmm o2 w. dd .5 _m ma da oz
_.w dd v; dd wd ~.a ad dé v; ed _.a Wm 0N ..m D
Ya dd ad ad ad Nd adm a.m ad a: a.a_ Wm ~.a _.m DZ :Eo>O
w._ dd vd dd ad _.m ad_ ad vd vd d.a ad _.m w; ma<
03 dd ON dd ad m.m_ d.a Na mN md _.2 ma NN mé m<OD 0
ad dd vd dd md Nd mam md vd _.m d.a_ d._ 0.2 we m<OQma<
m._ dd md dd dd v6 Nam ad nd ad ww— v._ ..a 9m ma< 02 099 emmm
ddm .d M: dd a._ ad m.mm d.d_ w; d._ :N _.n ma d.a m<OD
ad dd vd dd md mé mam wd vd _.m d.a_ d._ dd— wé m<09ma<
c; dd vd ad ad dd v.a_ dd vd md m.d_ 2 a4. ON ma< :Eo>O
0.2 dd Nu dd m._ a.: a: md ad ad ﬁn. Om Nn vh m<OQ
md. dd d.~ _d dd mé wd ..m dd md dd _.m Wm a.N ..d.
Nd. dd wd dd ad a.a _.N_ _.m md md a.a NM Wm ma ..E D
a.m dd a._ dd d._ dd. a.n a._ a._ md wd_ ma d.m ..m ..w
«a dd _._ dd ad ad v.3 a._ : vd vém M: N: ca :2 305.25
«..E dd ad ad ..m we :2 a.n ad ad 0: _.m ..a a.v ..w DZ ommm
md. dd dd _d dd m6 wd ..m d.m md dd 2” ON a.~ ..E
do dd ad dd wd Md adm _.m ad, vd 3: Ya ..w a4. ..E :Eo>O
vd dd : md wan _.a NE N6 _._ a._ «.2 v5 d.n dd ..w
I. dd ad Md 3 m.» adm _.~ wd ed 03 ed ad a.m ..a
v.: dd 2 ad ad Nd ad: ad 2 3 Wm— v.m ..c 0v ..v 32225 <2:
0 m U m U m U “— mZQ "5 m2? E5 0 m
Q75 "5 “—23 "S, =Eo>O comcquoU £93m 5:30
2.3 22:53 am odﬁwnsm am

econ was $833 am

 

 

 

 

 

 

 

wagons 8.5553 you 5 no 982: mo

88% N3 2an

250

The effects of design factors on longitudinal cracking—WP, based on this analysis, are
presented below:

HMA thickness: The effect of HMA thickness on longitudinal cracking-WP is
inconclusive. Sections with 4” HMA surface layer and sections with 7” HMA surface
layer have shown comparable levels of longitudinal cracking-WP.

Base thickness: The effect of base thickness on longitudinal cracking-WP is inconclusive.
In general, all sections have shown comparable performance.

Base Type: The effect of base type on longitudinal cracking-WP is inconclusive.
However, on average, sections built on ATB have exhibited least cracking compared to
other sections, especially among sections built on ﬁne-grained soils.

Drainage: In general, the effect of drainage is statistically signiﬁcant with un-drained
sections showing higher cracking than those with drainage. However, this effect is not
operationally signiﬁcant. This effect is statistically and operationally signiﬁcant among
sections built on ﬁne-grained soils, especially among sections located in WF zone. In
addition, drainage seems to be more effective for sections with DGAB.

On the whole, at this point in time, sections in WF zone have exhibited much
higher cracking than those in other climatic zones. Among pavements built on ﬁne-
grained soils, those built with DGAB have shown higher longitudinal cracking-WP and
those built with ATB have shown the least longitudinal cracking-WP. This main effect of
base type was statistically and operationally signiﬁcance. Also among pavements built
on ﬁne-grained soils, drainage has a signiﬁcant effect on longitudinal cracking and this
effect is more pronounced (signiﬁcant) among pavements built with DGAB. This effect

is statistically signiﬁcant and is of practical importance.

251

Mean std. deviate

Mean std. deviate

Maan std. deviate

1.20
0.70
0.20
--0.30
-0.80
-1.30

1.20
0.70
0.20
-0.30
-0.80

-l.30

1.20
0.70
0.20
-0.30
-O.80

-l.30

 

 

 

 

=v % "‘6 ans 2% e e e
.3? soiefe 8<§§§ E
OON
.. H <
DesignFactors <5
(a)Overall

 

l

     
 

I

C] Fine I Coarse

III

[IFII]

L.
..
...
..
..
..
,.
.-
.—
,..

 

 

 

4H

7"

8" (ND)
12" (ND)
8" (D)
12"(0)
16"(D)
DGAB
ATB
ATB/DGAB
PATB/DGAB
ATB/PATB
ND

D

Design Factors
(b) By subgrade type

 

DWF IWNF

 

 

 

4a
7"

8" (ND)

12" (ND)

8" (D)
12"(D)
16"(D)
DGAB

ATB
ATB/DGAB
PATB/[XEAB
ATB/PATB
ND

D

Design Factors

(c) By zone type
Figure 6-26 Effect of design factors on longitudinal cracking-WP

252

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

.oocudccoo do :32 3:E: .8 $3 “a anEcmG 98:35.... 265 .53 nudmnm 5qu
an an 2 2 E a 8 3. a we «2 z
3% 23 2:3 22 83 83 82 82 on: 83 as_eo.oz.m:m§eo m:
22 23 one 23 88 cm; on; aEEQaZaZMEEQ m<oo amigo
$3 83 83 83 82 22 UMEEaZaZwEeo =eu>o
maﬁa: .2 83m
$2 a; Mano $2 $3 82 £3 . Eéméoeméoaz
.2 E< .2 m<oo =<
89o a3 58 $3 :23 ES 32 82 $3 $3 “5238 o 85
8mm
. . . . . . . . . aOgmhts
£2 52 8; 33 a8 8: $3 83 So .eEﬁeméo oz
. . . . . . . . . ﬁgp—kx
98° 83 once 82 at; 82 23 82 $2 .:E<.:m<8 =eo>o
as $2 55 2:8 $2 $3 $2 82. $3 88 a8 ..Eaféafa o
22 :3 $2 $3 $3. $2 933 $3 2.2 88 82 5.2..” oz awmms
83 ES 32 S2 83 and 83 83 33 RS 53 ..£.m>..§.m>..m .355
83 3.. 2:8 ass 22 RS 82 on: 83 82 ES :25». 305.0222:
o T i a o a o a
"in no a2? a3 3.80 2.:
h*ZQ m0 m2? E5 =Eo>O cemtmanU ..ouumm :wﬁon.
ocou was odﬁwdsm am
23 0:25: a samba 3
AS»

-mEu—omco EEdBmmnoA '3?“de 935% do mnemtmmﬁoo How mos—ash mo Suﬁsm mad 033.

253

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

no» 3. .3 no 3. 2 0R n.9, 2. 3 <2 52 a: ”.2 o
92 3. 3 ..o 3 2 SN 2: 3 3 SN 3 ..E 2: oz m:
mg 3 to So 2 ..m 32.. was. to 2 in we a: m: o
. . . . . . . . . . . . . . o<oo $265
w: Cm mm 3 mm 2 a: 3m 3 mm mm». 3 IN 3: oz
NS 3 to MS «a an new a? 3 .2 «.8 0.2 n: 02 o =so>o
9: a: 2 am 3 3 9: N: 2 3 SN 3 9mm 5: oz
new od 3 S ..N 3 3m 2:. mo 3 <2 a2 a: ”S o:
as” 3. v.0 to am 3 3m at. as .2 EN n: N: NS 3.8 o
as ”a 3 S 3 M: n? «.2 3 mm mom «.2 3m ”.2 $85?
«.2 1m 3 _.o E 2 SN 2: 3 3 IN S a: 2: o: oz aboam
a? mom on m: on 3 Sc “.3 3 NM m2 3 EN a2 o<oo
as 2 a. S 3 3 3v «.2 a. 3. 0.2 2: NS ”.2 m<oooe<
a; No 3 3 mm an 3m 3m 3 3 we 2: 3: ed my: .355
3.5 n: 3. 2 an S can as. 3 2 SN a: ”.2 n: m<oo
NR ad no no 2 2 3m 9:. 9o 2 2m 3: S: n: ..E
9% od 3 2. am 2: E n: 3 2 98 3; NE a: ..E o
a: we 3 M: 3 2 EN n? I M: 2: ad 92 ad ..w
a: a2 2 9m 5 3 3m 3m 2 3 3m 3 3m oi ..E
. . . . . . oz “852533
3% a: 3 3 3 S Em 3m 2 3 3m 2 3m 3; ..w
NS go no 2. am 2 2m 9% 9o 2 SN 2; a: n: ..E
in 3 mm 2 am 3 2m Q? mm 3 SA. M2: 2: n: ..E =eo>o
30 Md 5 3 nm 3 Sun a: 5 3 3m 2: 3: a: ..w
«.8 3v 3 do 5 an 3m v.3. 3 2 3m 0: a: 3: ..a
. . . . . . . . . . . . . . “85.0232:
v8 a2 oo mm mm 2 2:. gm mo mm 2m N: m: a2 ..v
o o o a o o o o ..ozo no oz? ..E cameo 2.:
uzo do “23 “S, SE, cataeo 228a cased
o=o~ doe gnome; am was “ESE—o \E owﬁmnzm am = O 0
III!

 

 

 

 

Easiness 3:62:32 do.“ E do 3on do bmﬁesm 3.0 033i

254

Longitudinal Cracking- NWP
The effects of the design and site factors, in terms of standard deviate, are shown
in Figure 6-27. The summary of p-values corresponding to the analyses performed to
study the effects of design factors on longitudinal cracking-NWP is presented in Table
6-35. The mean PI corresponding to each comparison presented in Table 6-35 is shown in
Table 6-36. The effects of design factors on longitudinal cracking-NWP, based on this
analysis, are presented below:
HMA thickness: The effect of HMA surface thickness is not signiﬁcant. Comparable
amount of cracking occurred in sections with “thin” and “thick” HMA surface.
Base thickness: The effect of base thickness is not signiﬁcant. On average, sections with
16” base have shown slightly lesser cracking than other sections.
Base T ype: The effect of base type is not signiﬁcant. Comparable amount of cracking
occurred in all sections, irrespective of base type.
Drain age: In general, sections with un-drained sections showing higher cracking than
those with drainage. However, this effect is not signiﬁcant. Also, this effect is more
apparent among sections located in WF zone.
On the whole, at this point in time, it seems that longitudinal cracking-NWP is not
a “structural” distress. It may be more affected by climate. It may be noted that the

3310th of longitudinal cracking-NWP is higher among sections located in “freeze”

climate,

255

Mean Std. deviate Mean std. deviate

Man std. deviate

0.6

 

 

 

 

 

 

 

 

 

 

F-
l2
0.4 E
0.2 :
0.0 W - I T r I
I
-0.4 E
-0.6 E_ _
.. e a 6 e: e e e a e a o
<
DesignFactors < E
(a) Overall
0.6 _
E U ' I
GAE Fine Coarse
0.2 E .
0.0
-O.2 E
-0.4 1.
t l
-O.6 r E J
z, a " A o a a Q
22 fr? sééé. %
: : °° g :2 8 E
°° Q
<
DesignFactors 4' E
(b) By subgrade type
0.6 _
0.4
0.2 C
E
0.0 :
-o.2
-o.4 P
-O.6

 

 

 

 

 

4

70.

3" (ND)
12" (ND)
8" (D)
12"(D)
16"(D)

3

Design Factors
(c) By zone type

53
<

PATB/MAB
ATB/PATB
ND

D

Figure 6-27 Effect of design factors on longitudinal cracking-NWP

256

.uocuuccoomo .32 SnmE 8 $8 “a EmucEﬂm bﬂaozmzﬁm 32?. £8 393m ”802

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

on E 3 e. «a E 8 E ”S 2 a: 2
on; 23 SS 23 2.3 8% 83 cad uma_eoaz.m>ow2_so m:
236 82 a; SS 23 $2 32 82 89° 83 came um2_eo-oz.m>&2_eo m<oo $235
23 $2 83 82 2:3 33 83 33 ”maseoazéuwaia =eu>o
. . mP<QmP< .m> m.~.<n=m<0n_ mummm
:2 $3 $3 32 was 53 $3 $2 $3 22 .:Mwmmwwmmwm =<
:3 33 ES 83 9:6 $3 83 a; 58 $2 83 ”5.238 a 333%
82 33 80.0 $3 53 83 82 83 Ed E2 .:Eﬁwmﬁm oz
83 22. 993 82 83 82 83. owed 83 SS .:Emmﬁawm =eo>o
lass 33.0 33 £3 23 £2 and :3 22 SE ,.£.m>..m_.m>=w a
32 N23 83 :3 9.2 $2 53 23 23 She Ea?» oz mmwmwmé
ammo 8g 83 on: 82 82 22. 093 82 $2 82 ..E.m>..m_.m>..w =§>o
83 Ste 33 82 Sec 83 card 38 033 23 32 Lafv 305.0222:
o _ a o m o a o a "Ea "5 “EB Es 358 2E
mZQ ..E "~23 m3 =Eo>O camcmanU .22»; 5480
Eager 20:25:: ”sagas
/i

 

 

 

A32-m:.£om.5 ﬁﬁwammnoq lmBaFou “565% we mnemtmmﬁoo Sm 3:15-30 beﬁﬁsm mmb 2an

257

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

m; o.o ”on N. _ N o.: memo own o.o o.o“. W: o.:. ..mw m.mm ..oe D
o.o a; NS No 5mm ﬁeo mow od Now NS o.ow Mam 5% _.mm DZ mh<
_.o _.e 2:. o._m 9m: ego ﬂow o.o 3:. w.o_ e.e~. Non oSm Moe Q n
. . . . . . . . . . . . . . m<OD oonEmeQ
>3 mm o; moo e2 :3 mom of o; _: a; mum mom one DZ
NH o.m 0.: o._N NE o.mo o.ow Om 0.: o.o. fmn 5mm e.mm NSe Q :So>O
Q: oN 5mm 5o N6. 53 ooh es 5mm Q: w.~w odm o._e o.oe DZ
mg o.o ”.2. NAN Q: 98 o.mm o.o won 2.— o.:. ﬁmm w.mm ..oe mh<
o.o _.e no» o. _N 92 e._o ﬂow o.o Own o.o. e.e> mom osm m.we m<OQ Q
o.o m._ o.mw 02 mo. 53 ~60 no odw m.~_ mob wen o.om Moe m<OQ\m.r<
o.o M: Nam N.m ﬁmm hem mow oh NS NS oow m.mm 5% ..mm mh< CZ 099 025
5mm Wm 95 mo— e.m_ 53 won 92 o.o» ﬂ: o._w NSm w.wm oSe m<OD
o.o m; own 02 mo— 58 NS no ohm GS mob oem o.om Moe m<OQ\m._.<
he so Now w.e_ NE eﬁo ﬁne Em mow e.m_ QR. e.em o.oe owe mEr< :Eo>O
Num— oe mob N: oe— meo Own ~.o_ won 02 oi. o.om men wse m<OD
_.o o.o 3:. 92 m: meo 9mm o.o of. <2 mob m.mm o.mm ewe ..E
On— o.o 5E. mom 3: wow o.e\. ﬁn v.2. 0.3 mow own eSe NNm ..m_ D
o.o ﬁe 0.2. mom es moo o.om _.m 93. NS boo 5.5m e.om o.me ..w
es. ad ..wn ed e.o_ o.oo :o _.o~ ..wn o.o oi. oem o.wm o.oe ..m_
. . . . . . . . . . . . DZ 3056:: 35
mo om mow oo oo_ weo m; we mow o2 oww m.mm use e.mm ..w
_.o o.o of. 0.2 0: meo 9mm o.o 0.: of mom 3% ohm ewe ..E
meo— w; o.: o.m_ e.m_ ado moo _.o o5. 5.2 _.on ﬁmm woe o.oe ..m. :Eo>O
o.o me New :: o.e_ odo Nmo o.e New ”.2 o.on eom eSM o.me ..m
as no ﬂew of 5.2 woo o.oo me ﬂew 0.2 565 oem woe o.oe ..N.
. . . . . . . . . . . . . . mmoconH <2:
mo _e N: 52 n2 who oom no NE. o3 owb wmm omm use ..e
U n. U m U m U m "—20 m0 m2? m3 3300 02E
mZQ m0 m2? m3 :Eo>O acmtmoEoU 808mm swimm—
ocow 285:0 xm 03.33 Am

BEN 28 ooﬂwnam xm

 

 

 

 

 

 

 

mazocﬂoﬁo Ecmoammco— Mom E mo £82: .«o began—m omo oBmH

258

6.6.4 Effect of Experimental Factors on Pavement Response

This section of the report is a discussion of the results from analyses of FWD data
(pavement response) of the SPS-1 sections. Three parameters were chosen for
analyses — peak deﬂection under FWD load (do), far-sensor deﬂection ((16), and AREA.
The peak deﬂection under FWD load is indicative of the “overall capacity” of the
pavement structure while the far-sensor deﬂection is illustrative of the subgrade
“strength”. The AREA is the area under the ﬁrst three feet of the deﬂection basin. The
computational details regarding the AREA can be found in the reference “LTPP Data
Analysis: Feasibility of Using FWD Deﬂection Data to Characterize Pavement
Construction Quality”, NCHRP Project 20-50 [8], by Richard N. Stubstad, October 2002.
The AREA is indicative of stiffness of the upper layers of the pavement relative to the
stifﬁiess of the underlying layers. Higher the AREA higher is the stiffness of upper layers
in relation to underlying layers.

An ANOVA was conducted with the peak deﬂection under the FWD load plate
(do), the far sensor deﬂection at 60 inches from the FWD load (d6) and the AREA as the
dependent variables. All the response parameters have been calculated using the initial
deﬂections of the test sections.

It should be noted that the pavement surface temperature at the time of testing
was taken as a covariate along with the age at the time of testing and variability in the
HMA and base layer thicknesses. The natural logarithmic transformation has been
applied to the three response indicators to fulﬁll the ANOVA assumptions. The results
from ANOVA are summarized in Tables 6-37 and 6-38. The brief discussion of the

results is given below:

259

PeakDeﬂection under FWD Load (do)

When only design factors were considered by blocking the site effects,
interactions between HMA thickness and base type (p=0.043), base thickness and base
type (p=0.000), base type and drainage (p=0.000), have shown signiﬁcant effects on the
peak deﬂection.

Among the pavement sections built on DGAB, those with 4-inch HMA thickness
have shown higher peak deﬂection than those with 7-inch HMA thickness. Also as
expected, thicker bases for each base type have helped in reducing the peak deﬂection.
However, the reduction in peak deﬂection was more signiﬁcant in the case of ATB and
ATB/DGAB base types. Furthermore, among pavement sections built on DGAB, those
with drainage have shown lesser peak deﬂections than those without drainage. As
mentioned before, DGAB with drainage refers to PATB/DGAB with drainage. It was
assumed in the experiment, to study the effect of drainage among sections with DGAB
that PATB has the same structural “strength” as DGAB. This assumption appears to be
invalid from above result.

When all the site and design factors were considered simultaneously along with
the two-way interactions between the main factors, the interaction between subgrade
soil and climatic zone (p=0.005) was found to have a very signiﬁcant effect on the peak
deﬂection.

Among the pavement sections located in “wet” climates, those built on ﬁne-
grained subgrade soils have a signiﬁcantly higher peak deﬂection (do) as compared to
those built on coarse-grained subgrade soils. This effect is more prominent on

‘ pavements located in WNF zone than those located in WF zone.

260

Far Sensor Deﬂection ldg)

When only design factors were considered by blocking the site effects, the main
effects of base type (p=0.000), base thickness (p=0.005) and drainage (p=0.012) have
signiﬁcant effects on the far-sensor deﬂection (60-inches away from the center of the
load).

Pavement sections built on DGAB bases have shown higher far-sensor deﬂections
than those built on other base types. Pavement sections constructed on 8-inch bases have
also shown signiﬁcantly higher far-sensor deﬂections than those built on 12- or l6-inch
bases. Furthermore, sections built with drainage have lesser far-sensor deﬂections than
those without drainage. This effect can be attributed to PATB as the effect of drainage
and the effect of PATB cannot be separated (confounded).

When all the site and design factors were considered simultaneously along with
the two-way interactions between the main factors, an interaction between subgrade
type and climatic zone (p=0.000) was found to have a signiﬁcant effect on the far-
sensor deﬂections. Among pavement sections located in “wet” climate, those built on
ﬁne-grained subgrade soils have higher deﬂections than those built on coarse-grained
soils. This effect is signiﬁcant among sections located in WNF zone.

The ANOVA results also show that HMA thickness and pavement mid depth
temperature do not have a signiﬁcant affect on the far-sensor deﬂection. The results
seem reasonable, as this deﬂection (d6) represent the subgrade strength, which is

independent of the HMA thickness and pavement temperature.

261

alga

When only design factors are considered by blocking the site effects, the
interactions between HMA thickness and base type (p=0.002), base thickness and base
type (p=0.03), and, drainage and base type (p=0.000) have shown signiﬁcant effects on
AREA.

Among pavement sections built on DGAB, those with “thin” HMA surface layer
have signiﬁcantly lower AREA values compared to those with “thick” HMA surface
layer, implying that the upper layers of these pavement sections are “less stiff”. Also, the
increase in HMA thickness from 4 to 7 inches on ATB does not signiﬁcantly increase
AREA.

For sections built on DGAB, increasing base thickness from 8 to 12 inches has
not shown a signiﬁcant effect on AREA; however a two-fold increase in base thickness
(from 8 tol6 inch) has shown a signiﬁcant increase in AREA. Also, base thickness does
not seem to have a signiﬁcant effect on AREA in pavement sections with ATB bases.

Among the pavement sections constructed on DGAB, those with drainage have a
signiﬁcantly different AREA compared to those without drainage; test sections with
drainage have higher AREA, implying higher stiffness. This indicates that the structural
capacity of the PATB layer is somewhat higher than that of the DGAB.

When all the site and design factors were considered simultaneously along with
the two-way interactions between the main factors, the interaction between subgrade type
and climatic zone (p=0.000)iwas found to have a very signiﬁcant effect on AREA.
Among the pavement sections located in WNF zone, those built on ﬁne-grained subgrade

soils have signiﬁcantly higher AREA values than those built on coarse—grained soils.

262

However, in the case of sections located in WF zone, this effect is not signiﬁcant

indicating that AREA could be independent of the subgrade soil type.

Table 6-37 Summary of p-values from ANOVA for determining the effect of design
factors on ﬂexible pavement response — Overall

 

 

 

 

 

 

 

 

 

 

 

 

 

Performance Measures
Desi Factor
gn Peak Deﬂection (do) Far Deﬂection (d6) AREA
HMA thickness 0.000 0.560 0.000
Base type 0.000 0.000 0.000
Base thickness 0.000 0.005 0.214
Drainage 0.590 0.012 0.000
M‘“ depth 0.000 0.733 0.000
temperature
Site (blocked) 0.000 0.000 0.000
R2=0.884 R2=0.864 R2=O.854
N=210 N=210 N=210

 

Table 6-38 Summary of p-values from AN OVA for determining the effect of site factors

on ﬂexible pavement response —— Overall

 

 

 

 

 

 

 

 

 

 

 

Performance Measures
D ' F t

esrgn ac or Peak Deﬂection (do) Far Deﬂection (d6) AREA
Subgrade 0.000 0.000 0.353
Zone 0.000 0.000 0.000
Subgrade*Zone 0.005 0.005 0.005
M“ depth 0.000 0.495 0.000

temﬁrature

R2=0.865 RT=O.658 R2=O.682

N=210 N=210 N=210

 

263

 

 

6.7 APPARENT RELATIONSHIP BETWEEN RESPONSE AND
PERFORMANCE

In this section of the report the observations regarding apparent relationships
between ﬂexible pavement response (FWD testing) and performance are presented. The

usefulness of such relationships can be divided into two categories:

- Explanatory: To provide an explanatory information for a given performance
trend. For example, a relationship between AC rutting and the farthest sensor
deﬂection would indicate that rutting is related to the subgrade soil.

- Predictive: To provide a predictive capability of the future level of a given
performance measure. For example, the initial high average deﬂection of a
section may explain its future cracking and rutting (due to subgrade)
performance.

Explanatory relationships were established using multiple regressions on data
from all the test sections in the experiment. Predictive relationships were established
based on bivariate correlation analyses at the site level, and using scatter plots on data
from all sections. The DLR data were used for predictive relationships regarding the

instrumented sections in Ohio.

6. 7. 1 Overall Analysis—Explanatory Relationships

In this section, the entire p0pu1ation of the SPS-1 experiment was used to seek
apparent explanatory relationships between response and performance. This analysis was
done irrespective of the experimental design matrix layout, since pavement response
should reﬂect the effects of the various structural designs. In other words, the analysis
spans over all the SPS-l sections, as opposed to it being restricted to individual structural
designs. The spatial variability of the deﬂections and deﬂection-based indices (within a

section) was considered by taking the 95th percentile within each section. As deﬂection

264

on all sections was measured during different seasons and times, the impact of
temperature and moisture conditions cannot be ignored. Additionally the deﬂection and
deﬂection-based parameters (SCI, BDI etc.,) are inﬂuenced by variety of factors, such as:

Asphalt temperature (at mid depth)

Thickness of asphalt layer

The layer moduli of various layers and overall pavement structure
Subgrade strength

Apparent stiff layer depth

Pavement distresses etc.

To consider the effect of various variables on the response at the same time, the
multiple linear regression technique was used. The pavement response parameters
(surface deﬂection (do), SCI and BDI) were taken as dependent variables and all other
variables (temperature, asphalt thickness, subgrade strength and pavement distresses)
were considered as independent variables. As expected, the surface deﬂection is
. signiﬁcantly correlated with the asphalt layer thickness, mid-depth asphalt temperature,
and deﬂection at the outer most sensors that represents the subgrade strength.
Furthermore, fatigue cracking, longitudinal cracking—WP and transverse cracking in all
the sections have shown statistically signiﬁcant relationships with the surface deﬂection.
The results of multiple regression analyses within each zone have also shown that, on
average, fatigue and transverse cracking has a signiﬁcant positive effect on the surface
deﬂection. An example of such multiple regression models is given by the following

equation:

1n(d0) = 3.694 + 0.17T — 0.0851106 — 0.03 lAge +
0.6021n(d6) +0.001AC+0.004TC+0.001LCWP

(R2=0.856, SE=0.269)
where: In is the natural logarithm
T is the mid-depth asphalt concrete temperature (C)

265

Hac is the HMA layer thickness

Age is the age of the pavement section at the time of FWD testing
AC is alligator cracking (sqm) at the time of testing

TC is transverse cracking length (m)

LCNWP is longitudinal cracking not in the wheel path (111)

The sensitivity analysis of the regression model for overall SPS-1 database was
performed to observe the explanatory relationships between various independent
variables with the surface deﬂection (do) under the FWD load plate. The following

conclusions can be made from these results:

a The effect of asphalt thickness on the measured surface deﬂection (do) is very
signiﬁcant (p=0.000). The thicker the HMA layer, the lower the deﬂection will be.

- Mid-depth asphalt temperature at the time of testing has a signiﬁcant positive effect
on do (p=0.000).

- The age of the pavement has indicated a negative effect on do (p=0.000). Aging
effect on HMA pavement may cause the stiffening of asphalt thus may reduce the
deﬂections.

o The higher the “subgrade” deﬂection, d6 (deﬂection at the outer most sensor, or 60
inches in this case), the higher do will be (p=0.000).

- Fatigue cracking (p=0.000), longitudinal cracking-WP (p=0.006) and longitudinal
cracking-NWP (p=0.012) have a signiﬁcant positive effect on do; i.e., higher
cracking will cause an increase in do.

0 Similarly, transverse cracking has a signiﬁcantly positive effect on do (p=0. 001).

6. 7.2 Site Level Analyses— Predictive Relationships

This section summarizes the ﬁndings regarding predictive relationships between

initial response (FWD deﬂection or deﬂection-based indices) and future pavement

266

performance (cracking, rutting and roughness), at the site level. Various deﬂection-based
indices [8, 9] were calculated based on the individual deﬂection basins for each section;

these indices include:

AREA (the area under ﬁrst three feet of deﬂection basin),

SCI- Surface Curvature Index, (d0 — d12),

BDI— Base Damage Index, [9](d12 — d24),

d36 - (d0-d36),

do (peak deﬂection under the load),

d6 (farthest deﬂection at 60 inches away from the load),

ES (effective stiffness of upper (bound) layer), and

Eg (subgrade modulus calculated from surface deﬂection at 36 inches from the load).

Bivariate correlation analyses between response parameters (deﬂections or
deﬂection basin parameters) and performance (cracking, rutting and roughness) were
conducted for all the states within SPS-1 experiment. The latest performance for each
section within the SPS-1 experiment was used in these analyses. The effect of
temperature on the measured deﬂection was taken into account by applying a temperature
correction [10] . It is to be noted that for a site age, trafﬁc, construction, material
properties and environment are the same and thus this provides a good opportunity for

seeking apparent relationships.

Figure 6-28 and Figure 6-29 are examples of bivariate relationships between SCI
and AREA with fatigue cracking. The site in the state of Kansas (20) was chosen for this
example because of high extent of cracking at the site. It can be seen that for the sections
in this site, initial SCI and initial AREA have a slight association (p = 0.4) with the future
fatigue cracking, in that higher the SCI or lower the AREA, higher is the cracking.

Similarly, Figure 6-31 is the relationship between BDI and future rutting for the same

267

site. In this case, BDI has a strong correlation (p = 0.77) with future rutting i.e. sections
that had higher initial BDI have higher rutting at a later stage. Also, Figure 6-30 shows
the variation in future roughness (IRI) as a function of BDI. In this case, BDI has a
correlation (p = 0.42) with the future roughness of all the pavement sections for this site
(KS (20)).

Figure 6-32 and Figure 6-33 show relationships of roughness and rut depth with
BDI and AREA for the sections in the site of OH (39). Strong correlations (p = 0.85
each) were observed between future roughness and rut depth. Sections that had higher
initial BDI had higher future roughness, and sections with lower AREA had higher future

rutting.

268

 

A 400
g 350 l y=l.7875x+126.76

:9 300 R2=0.1661 ,

3:.“ 250 °

‘53 200 . ,

63 150 .

a 100 °.

,3 50 1

L1- 0 l l i rum L I 4 any

 

 
 
 
  

 

 

SCI

Figure 6-28 Fatigue cracking and SCI relationship-

 

 

 

 

   

 

 

 

State (20) Kansas
3.0
2,5 y= 0.0096x+ 1.0706
A 2.0 R’=0.178 .
g 1.5 [ ° °
:2 1.0 . . ° . , °
0.5 t"
0.0 l 4 *
l 10 100
BDI
Figure 6-30 Roughness and BDI relationship—
State (20) Kansas
5.0 t;
y = 0.0176x + 1.4784
4.0 ~ ,
A R =0.7185
£5 3.0 —
V 2.0 ~
g o o
1.0 -
0.0 4 * 1 l
l 10 100 1000

BDI

Figure 6-32 Roughness and BDI relationship—

State (39) Ohio

 

 

 

 

 

 

 

 

 

 

 

 

 

A 400
g. 350 . y=-14.815x+548.98 .
if, 300 F 2-
R -—0.l43 o
.1
§ 200 i , a
U 150 o
3 100 - ."
DD
2;- 50 -
LL. 0 l I 1 L l 1 I
0 5 10 15 20 25 30
AREA
Figure 6-29 Fatigue cracking and AF relationship—
State (20) Kansas
35 F
A 30 t y=0.2283x+ 3.8001 .
E 25 t R2=O.6012
E 20 t
g 15 ~
0.» IO '—
e 5 .
0 ._
1 10 100
BDI
Figure 6-31 Rut depth and BDI relationship—
State (20) Kansas
25 l
A =-1.2739x+41.542
F: 20 l- y 2
g . R =o.7329
€- 15 l
5 10 g—
... l
52 5 t
0 L 1 I
10 15 20 25 30
AREA

Figure 6-33 Rut depth and AF relationship—

269

State (39) Ohio

Tables 6-39 through 6-41 are summaries of correlation coefﬁcients from the
bivariate analyses for three performances (fatigue cracking, rutting and roughness) and
various deﬂection parameters between all the sites.

The results show that ﬁfteen out of seventeen sites have a consistent trend of
relationship between AREA, BDI, do and future cracking. Also, fourteen out of seventeen
sites have a positive association between SCI and fatigue cracking. On an average,
AREA, SCI and BDI have reasonable associations with fatigue cracking for all the sites
in SPS-1 experiment. Sections that have higher fatigue cracking had higher initial SCI or
BDI, and lower AREA.

The deﬂection basin parameters do not have a consistent association with rutting
across the sites (see Table 6-40). This inconsistency may be explained in light of different
rutting mechanisms for ﬂexible pavements i.e., structural or asphalt mix rutting.

Consistent trends were observed only between EDI and future roughness across
most of the sites (15 out of 17 sites) in the SPS-l experiment (see Table 6-41). Sections
that had higher BDI have higher roughness.

Apparent relationships between AREA and various performance measures
(fatigue cracking, rutting and roughness) were found to be signiﬁcant within sites that
have shown considerable distress. Higher AREA means stiffer upper layers of a
pavement. Sections that had higher AREA exhibited lesser cracking, rutting and
roughness. Based on the magnitude of correlation coefﬁcients, it was also found that
sections that had stiffer bound layers are more likely to exhibit cracking than (structural)

rutting.

270

Table 6-39 Summary of correlations for deﬂections and DBPs with fatigue cracking

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

State Area ES/Esg ES Esg d0 d6 SCI BDI Zone 86
31 Nebraska -0.79 -0.66 -0.64 -0.37 0.90 0.43 0.94 0.91 WF F
26 Michigan 0.48 0.59 ‘ 0.42 0.21 -0.42 -0.03 -0.46 -0.45 WF F
19 Iowa -0.45 -0.1 l -0.09 0.05 0.41 0.15 0.55 0.49 WF F
20 Kansas -0.38 -0.26 -0.25 -0.14 0.33 -0.01 0.41 0.30 WF F
39 Ohio -0.66 -0.44 -0.44 -0.43 0.58 0.20 0.63 0.63 WF F
55 Wisconsin -0.46 -0.38 -0.37 -0.02 0.38 0.04 0.51 0.22 WF C
10 Delaware -0.93 -0.78 -0.65 -0.04 0.72 -0.03 0.93 0.73 WF C
5 Arkansas -0.07 -0.02 -0.19 -0.54 0.34 0.59 i 0.17 0.34 WF C
51 Virginia -0.72 '-0.57 -0.58 -0.32 0.75 0.04 0.79 . 0.75 WNF F
1 Alabama -0.79 -0.68 -0.64 -0.08 0.72 -0.27 0.75 0.73 WNF F
48 Texas -0.48 -0.33 -0.57 -0.49 0.74 0.65 0.78 0.58 WNF C
40 Oklahoma -0.47 -0.43 -0.59 -0. 16 0.25 -0.05 0.28 0.28 WNF C
12 Florida —0.40 -0.37 -0.40 0.14 0.50 -0.12 0.46 0.55 WNF C
30 Montana -0.36 -0.31 -0.58 -0.74 0.53 0.83 0.34 0.42 DF C
32 Nevada -0.49 -0.355 -0.31 0.10 0.38 -O.17 0.41 0.29 DF C
35 New Mexico 0.19 0.22 0.33 -0. 14 0.31 -0.01 -0.14 0.60 DNF F
4 Arizona -0.13 -0.22 0.06 0.53 -0.10 -0.55 -0.03 -0.03 DNF C
(-) p 15 15 14 12 2 9 3 2
(+) p 2 2 3 5 15 8 14 15
Mean 04] -0.30 -0.32 -0.14 0.43 0.10 0.43 0.43
Std 0.36 0.34 0.34 0.32 0.33 0.35 0.38 0.33
CoV 0.89 1.12 1.03 2.21 0.76 3.55 0.89 0.76

Note: The SPS-l sections in State 22 (Louisiana) are young and have not shown any signiﬁcant distress

therefore, are not included in this analysis.

271

 

Table 6-40 Summary of correlations for deﬂections and DBPs with rut depth

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

State Area ES/Esg ES Esg d0 d6 SCI BDI Zone SG
31 Nebraska -045 -O.48 -024 0.29 0.28 0.06 0.41 0.29 WF F
26 Michigan 0.32 0.53: 0.41 0.10 -0.14 -004 -0.16 -014 WP r
19 Iowa -0.56 -043 -041 -022 0.40 0.13 0.43 0.46 wr r
20 Kansas -0.80 -0.55 -059 -0.23 0.76 0.11 0.82 0.78 wr 1=
39 Ohio -0.86 -0.88 -0.88 -072 0.79 0.51" 0.75 0.76 WF F
55 Wisconsin 0.26 0.31 0.37 0.37 -0.60 -0.45 -0.48 -O.65 WF C
10 Delaware -072 -055 -0.62 -025 0.66 0.37 0.73 0.61 W? C
5 Arkansas 0.37 0.51 0.51 0.03 -011 -002 -020 -003 WP c
51 Virginia -0.58 -040 -039 -0.22 0.69 0.02 0.78 0.68 WNF F
1 Alabama -051 -034 -0.26 0.23 0.63 -030 "0.73. 0.69 WNF r
48 Texas 0.65 0.62 0.80 0.37 -O.67 -0.28 -0.65 -O.62 WNF c
40 Oklahoma 0.02 0.23 -015 -043 0.60 "0.67 0.54 0.55 WNF C
12 Florida 0.56 0.60 0.40 -0.62 -033 0.66 40.44 50.39 WNF C
30 Montana -0.62 -0.66 -070 -0.59 0.68 0.56 0.60 0.63 DF c
32 Nevada 0.05 -0002 -003 -031 0.03 0.13 —0.10 0.25 DF c
35 New Mexico 0.35 0.27 0.31 -0.24 -0.46 0.25 -047 -009 DNF F
4 Arizona -019 -0.31 -033 -020 0.06 0.02 0.05 0.04 DNF c
(-)p 9 10 11 ll 6 5 7 6
(+)p 8 7 6 6 ll 12 10 11
Mean _ -0.16 -009 -0.11 -0.16 0.19 0.14 0.20 0.22
Std 0.51 0.50 0.49 0.34 0.51 0.33 0.52 0.48
CoV 3.22 5.46 4.65 2.20 2.61 2.33 2.66 2.13

 

 

 

Note: The SPS-l sections in State 22 (Louisiana) are young and have not shown any signiﬁcant distress
therefore, are not included in this analysis.

272

Table 6-41 Summary of correlations for deﬂections and DBPs with IRI

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

State Area ES/Egg ES Esg d0 d6 SCI BDI Zone SG
31 Nebraska -0.61 -044 -0.54 -043 0.66 0.49 0.62 0.72 WF F
26 Michigan -075 -0.71 -0.78 -0.82 0.78 0.74 0.73 0.76 F
19 Iowa -053 -031 -0.28 0.04 0.18 -012 0.30 0.25 F
20 Kansas -034 -031 -0.38 -0.38 0.40 -0.16 0.27 0.42 F
39 Ohio -079 -0.65 -0.65 -0.58 0.84 0.37 0.85 V 0.85 F
55 Wisconsin 0.37 0.45 0.50 0.23 -054 -004 -0.46 -o.54 WF c
10 Delaware on -0.60 -0.68 -O.36 0.73 0.28 0.68 0.76 WF c
5 Arkansas -0.02 -0.03 -0.l7 -0.41 0.21 0.42 0.05 0.21 WF C
51 Virginia -070 -050 -055 038 0.78 0.06 0.83 0.77 WNF F
1 Alabama -047 -032 -031 -031 0.69 0.23 0.66 0.69 WNF F
48 Texas 0.38 0.44 0.36 -017 -011 0.15 -0.20 -009 WNF c
40 Oklahoma 0.45 0.49 0.25 -035 0.17 0.56 0.10 0.10 WNF c
12 Florida -O.31 -0.31 -042 0.14 0.47 -0.15 0.38 0.51 WNF C
30 Montana -0.55 -0.61 -O.69 -0.62 0.63 0.62 » 0.50 0.55 DF c
32 Nevada -0.47 -0229 -020 -0.16 0.71 0.44 40.72 .- 0.68 DF c
35 NewMexico 0.39 0.50 0.49 -031 -002 0.33 -035 0.29 DNF F
4 Arizona -0.56 -0.46 -0.47 -0.08 0.63 0.05 0.64 0.65 DNF C
(-)p 13 13 13 14 3 4 3 2
(+)p 4 4 4 3 14 13 14 15
Mea
n -031 -021 -027 -029 0.43 0.25 0.37 0.45
Std 0.44 0.42 0.42 0.27 0.39 0.28 0.41 0.37
CoV 1.44 2.00 1.57 0.93 0.91 1.12 1.11 0.83

 

 

Note: The SPS-1 sections in State 22 (Louisiana) are young and have not shown any signiﬁcant distress
therefore, are not included in this analysis.

273

6. 7.3 Overall Analyses— Predictive Relationships

This section summarizes the ﬁndings of apparent relationships between initial
response (FWD deﬂection or deﬂection base indices) and future pavement performance
(cracking, rutting and roughness), based on data from all the test sections in the SPS-1
experiment. The relationships were explored using bivariate scatter plots between
selected response parameters and performance measures for all the pavements in the
experiment.

Though the sites differ in age, trafﬁc, climate, and materials this analysis is
intended to use the wealth of data from all the sections in the experiment. Moreover, the
variation in age of the sites may not be very critical at this point in time as no deﬁnitive
trends were observed between pavement age and performance (see Figure 6-34). Also, it
is assumed in this analysis that deﬂection basin parameters (pavement response) will
“characterize” the structural features such as HMA surface thickness, base type and base
thickness. In other words, pavement response was assumed to be strongly correlated with
the structural capacity of the pavement. In order to account for the effects of subgrade
type and climate, relationships were explored for different subgrade soil types (ﬁne- and
coarse-grained soils) and climates (WF, WNF, DF and DNF).

Figure 6-35 shows a scatter plot between SCI (from initial response) and fatigue
cracking. Among pavements constructed on ﬁne-grained soils, ones with higher initial
SCI have more cracking, especially in WNF zone. Similarly, from Figure 6-36 it seems
that stiffer pavements (higher AREA) on ﬁne-grained soils, especially if located in WF

climatic zone, have higher fatigue cracking.

274

Higher longitudinal cracking-WP was observed for the pavement sections with higher
initial AREA, especially among pavements located in WNF climatic zone (see Figure
6-37). The observation may imply that pavements with stiffer structural layers are more
likely to exhibit this distress. No apparent relation was observed between AREA and
longitudinal cracking-NWP (see Figure 6-3 8). The distress was found to be independent
of the structural capacity (AREA) of various pavements. Thus the probable cause of this
distress type may be the environment and not the loading (trafﬁc).

Figure 6-39 shows a scatter plot between AREA and transverse cracking. No
apparent trend was observed in the plot. This could imply that this distress type is not
load-related and probably caused by the environment.

The apparent relationship between initial BDI and rutting is shown in Figure 6-40.
It seems that among pavements constructed on ﬁne-grained soils and located in WF zone,
those with higher BDI experienced higher rutting. It can also be observed that some
pavements with less BDI (stronger structure) have experienced higher rutting as
compared to pavements with high BDI. These pavements could have experienced mix-
related rutting (not structural rutting).

Figure 6-41 is the scatter plot between BDI and latest IRI. It seems that among
pavements constructed on ﬁne-grained soils and located in WF zone, those with higher

BDI developed higher roughness.

275

Fatigue Cracking (sq-m)

Long. Cracking-NWP (m)

Rut depth(mm

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

350<
SG - '
3m” . C .
250< ' F '
I
200. .' : .
1504 _ ' .
100- ' " '
I I I I
50- : I I. I ' .l
I I - I '
04 I .3. .J.‘ I r I,
0 2 4 6 8 To
A8=(years)
(3) Fatigue cracking
300 SC ' I
C I 3
250a ' 0
F I . I I
I. : I
XX): . . I I '
a. '
150. i I I I
I
I I I I
K”3 .: . '
Il --. -
501 I . . '
04 I I -Lliii I :-
0 2 4 6 8 I?
Ascb'eaIS)
(c) Long. cracking-NWP
30‘ . . ‘ Zorn
I u: .
25« ‘ o - our
A A : . WP
20" O . A WNF
o 3 ’ A
154 . g‘ £ ’ I I
E
10. o
. . gt 1
51 i; 2%28? I”;
o- . T
0 2 4 10
Aseoears)
(e) Rut depth

Transverse Cracking (m) long Cracking-WP (m)

IRI (tn/km)

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

250
so "
C I . -
ZW‘ F I
' I
150. ‘ I g I
i I
I I
la). ' I . I
I . '
504 . I = I
C _ I t '
04 I I. bl“ I
5 2 4 6 8 10
Age (years)
(b) Long. cracking-WP
90‘ SG
1 I
80 . C
70‘ I F I
60 '
I
sol . . I
40.
30‘ I I
zol
I I I . =
10‘ . '
0< I "I I hi. I. I I
0 2 4 6 8 10
A8: (ream)
((1) Transverse cracking
4.5 2“
4.0< ’ - Dr
- DNF
3.5‘ O . WP
3.04 ‘ WNF '
. O
2.5«
2.04 . 1 ‘ z . O .
’ CA I
1.54 0 .0.! "C. A
1.0- ‘ ’Q ' 0‘ c
0'5‘ 1 r I v I f
0 2 4 6 8 10
Age (years)
(i) Roughness

Figure 6-34 Relationships between age and different performance measures

276

 

 

 

Fatigue Cracking (sq-m)

Fatigue Cracking (sq-m)

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

350~
I SG
300‘ I I C
I - F
250- '
I
200~
150-
100-
50-
O A A 1 n
1000
(a) Effect by subgrade soil type
350-
’ A Zone
300- 0 DF
9
I DNF
250- t
o 9 WF
200- Q ‘ . ‘ A WNF
c. O
150 g ‘
A
o ’A
50’ ' .'?¢ 9- .
9 cu I 3 ’
0- "M-
10 100 1000
SCI

(b) Effect by climatic zone
Figure 6-35 Apparent relationships between SCI and fatigue cracking

277

Fatigue Cracking (sq-m)

Fatigue Cracking (sq-m)

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

350
I SG
300. ' 0 C
I . F
250. '
I
200- I
" I
. I
150 I ..
I
I
100‘ . : :I I
. I. I I.
50" . I.- . . ...I ‘I :II‘.
0- ' .- II :I—uiw'
1‘0 1'5 20 25 30
Area Factor
(a) Effect by subgrade soil type
350- .
Zone
300: ‘ . DF
0
250- t ' DNF
O 0 WF
200‘ 9 ﬂ 0 A WNF
- c
150 A . ..
A A '
100- o o I .6 .
A 9 I a 9 g
50" . . At I O a. .
.' 8' ’ . 9‘ z"?
0~ ' J."
10 15 20 25 30
Area Factor

(b) Effect by climatic zone

Figure 6-36 Apparent relationships between AREA and fatigue cracking

278

 

 

 

 

 

 

 

 

 

 

 

 

 

 
 

 

 

 

300 ~ SG ' I
o C ‘ . I
250' I F I = ' I.“ :
A f I I I I
g 200- . I I. I“
I w . I .'
E 150 - O. .3 l ‘ J ID:
I I I- } I -
B 100 ' : :.. I.
. I I... ...I I. I
50 - I ' l u
u 'u' 3' . u. '3' -
Or- . I II."I-.I. IM
10 15 20 25 30
Area Factor
(a) Effect by subgrade soil type
250 ' .
Zone . .
o I
200b DF .
I DNF .
A WF ' '
E 150- : WNF . o A. A
g I I I I
I I
u - O
U 100 Q ' I I .
»—l I .
50 ‘ IA
I ‘ I
0 . I III. “¢"
10 15 20 25 30
Area Factor

(b) Effect by climatic zone
Figure 6-37 Apparent relationships between AREA and longitudinal cracking-WP

279

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

300- SG ' I
' C i . I
250T I F I = I... :
I I D I
a 200'- f I I I.
v I ' . .II‘
E 150~ I. I ' q... I'I
I I I
. I I
I I .- I I I
B 100- I = :.. 'I
. I I ...l.. ~ .
_ I
50 I. =. I: .‘II .
- I
0- .III II.” . -M
10 15 20 25 30
Area Factor
(a) Effect by subgrade soil type
300- Zone " A
. DF % . 9
_ I
250 ' DNF O o o
A 9 WF ‘9 . z I . ~ . I
8 200‘ . WNF o .A ”
E 150- 3A a - W J «I.
9" A ' O '
l I * AA.
0 100- x .A A
. A u A I“ a.
50~ ‘A
I l. A A
0» '.l #3.:- AM
10 15 30
Area Factor0
(b) Effect by climatic zone

Figure 6-38 Apparent relationships between AREA and longitudinal cracking-NWP

280

 

Transverse Cracking (m)

Transverse Cracking (m)

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

90-
SG
80- . C C
70" I F I
60" I
I
50- .. I
40t
I
30- I '
I
20- . . I'- ll I I
10- ' " .' ﬂ? '
.. .I v. .
0- ..I- a I
10 15 20 25 30
AreaFactor
(a) Effect by subgrade soil type
90t
Zone
80b . DF '
70" I DNF .
60- . WF ’
I
_ A WNF
50 A. .
40-
O
30- I 0'
20 ’ '
' A 0 It
I t I ‘
10_ I I. . 'g Q
I. CA w I
0- III‘ I
10 15 20 25 30
AreaFactor

Figure 6-39 Apparent relationships between AREA and transverse cracking

(b) Effect by climatic zone

281

 

 

 

 

 

 

 

”100

l ‘10

 

 

A83 ﬁg 3

BDI '

(3) Effect by subgrade soil type

 

 

 

 

 

 

 

100

 

 

gees saw 3

10

BDI

(b) Effect by climatic zone

Figure 6-40 Apparent relationships between BDI and rut depth

282

IRI (m/km)

IRI (In/km)

 

 

 

 

 

 

 

 

 

 

 

 

 

 

4.5-
so .
4.0~ . c
3.5- I F
3.0- I I I
2.5~
I I
2.0” . I I i I I
I. .- ~ # I I
1'5- f. ..E. } I I
r-mw- '
0.5- H 1 l i
1 110 100
BDI
(a) Effect by subgrade soil type
45.
Zone
4.0~ . DF "
I DNF
3.5- . WP
3.0- ‘ WNF ‘
. O
2.5-
2.0-
1.5-
1.0-
I
0.5 3
l

BDI
(b) Effect by climatic zone

Figure 6-41 Apparent relationships between BDI and IRI

283

 

 

 

6. 7.4 Dynamic Load Response for 0H (39) test sections

This section presents the summary of ﬁndings from the analysis of Dynamic Load
Response (DLR) data from the instrumented ﬂexible pavement sections in the state of
Ohio. These sections were instrumented with strain gauges, pressure cells and LVDTs to
measure the pavement “response”. The SHRP experiment in Ohio targeted four core
sections (see Chapter 2) for the installation of sensors to monitor dynamic pavement
response during controlled vehicle tests.

The main objective in this project was to study the response-performance
relationship by using the measured dynamic load response and actual observed
performance of the sections, in the SPS-1 experiment. Therefore, an attempt was made to '
relate the observed performance of these instrumented sections with measured responses
(strains and surface deflections in HMA surface layer, and stress at top of subgrade) by
means of bivariate scatter plots.

The bivariate relationships between measured responses (tensile strain at bottom
of asphalt layer, compressive stress at top of subgrade and surface deﬂection) and
observed performances (fatigue cracking and rutting) are shown in Figure 6-42. Also, the
bivariate relationship between response (compressive stress at tope of subgrade and
surface deﬂection) and observed performance (roughness in terms of IRI) is shown in
Figure 6-43. However, these ﬁndings are limited to these four instrumented sections. The
measured longitudinal strain (initial value) is “strongly” associated with future fatigue
cracking, and the vertical stress at the top of the subgrade is “strongly” associated with
future rutting. Other observations regarding the dynamic load response of the

instrumented test sections are summarized below:

284

In general, the strain in the longitudinal direction is higher than the strain in the
transverse direction; this is consistent with the mechanistic analysis for ﬂexible
pavements.

Sections with higher strain values have poor fatigue performance. These results are
in agreement with the mechanistic-empirical design predictions as fatigue cracking
in the ﬂexible pavements is generally considered to be related to the initial tensile
strain at the bottom of the HMA layer (bottom up cracking).

The sections that exhibited high measured stress at the top of the subgrade and high
surface deﬂection have shown poor rut performance.

The sections that exhibited high measured stress and deﬂection have higher

roughness.

285

 

 

Fatigue Cracking (sqm)

N
(I!
h I" “T—‘r'T' I"‘-|'—F'—I' r—f_‘T ”I 'I—1

0 50 100 150 200 250

Long. Strains (micro-strains)
(a) Relationship between strain and fatigue cracking
16
14

12
10

 

y = 0.6.r1'15

  

Rut Depth (mm)

 

*T'”! r—r—v 1"T—r-I i—T-r 'I“l—1‘"l

ON-D-OOO

0
£11

10 15 20
Stress at top of subgrade (psi)
(b) Relationship between stress and rutting

 

16
14
12
10

I-T.Y"I—T—‘

' T

Rut Depth (mm)

'-|" I inf—I"

 

1" l l l

ONAOOO

L L l 4 i I L ' 1 i 1 . i

0 20 4O 6O 80

Surface Deﬂection (mils)
(0) Relationship between deﬂection and rutting

Figure 6-42 Relationship between measured responses and observed performances—-
Fatigue cracking and rutting

286

 

 

 

2.5 i
2.0 ~ 1
_ ' i

A I
E 1.5 L 033‘
g I y: 0.77X . i
E 1-0 ” R2==O.7164 l
0.5 — '
l
0.0 e L 1 ‘
0 5 10 15 20

Stress at top of subgrade (psi)
(d) Relationship between stress and roughness

 

 

 

 

2.5 [
F O
2.0 - .
Q h I
g 1.5 r .
g 1.0 _ y:1.1x0.156
2.—
0.5 _ R f0.4399
00 a i 1 r i J i #1 i i L
0 20 4O 60 80

Surface Deﬂection (mils)

(e) Relationship between deﬂection and roughness

Figure 6-43 Relationship between measured responses and observed performances—
Roughness

287

6.8 SYNTHESIS OF RESULTS FROM ANALYSES

This section of the report summarizes the findings from various analyses
performed on SPS-l data. The methods employed in this study were explained in Chapter
4 and the results obtained from these analyses were presented above in this chapter.

Broadly two types of analyses were employed — magnitude-based and frequency-
based. The magnitude-based analyses that were used are one-way (univariate) and
multivariate AN OVA. These methods are used for comparison of means. The frequency-
based analyses that were used are Binary Logistic Regression (BLR) and Linear
Discriminant Analysis (LDA). These methods help identify the factors that signiﬁcantly
contribute to the occurrence of a distress based on the likelihood of occurrence or non-
occurrence of distresses. In site-level analyses, the performance of pavements withineach
site was compared. The results from site-level analysis were used to ascertain the
consistency of the effects of experimental factors across all sites.

The magnitude-based methods, though powerful, are more appropriate for
analyses of distresses which have both high occurrence and magnitude (for example:
fatigue cracking, roughness, and rutting). On the other hand, the frequency-based
methods are more suitable when the occurrence of a distress is fairly high (for example:
transverse cracking) but magnitude is low.

An attempt has been made to summarize the above said effects of design and site
features on the performance and response measures. The results were interpreted in light
of the type of analysis, and occurrence and extent of distress. ANOVA being the most
“powerful” among the methods was given higher importance. However, the results from

this analysis may not be reliable in case of limited (low occurrence of distress) or

288

unbalanced data. Therefore, in these cases, the effects of design features, on the
occurrence of distresses were investigated using BLR and LDA.

All results need to be interpreted in light of the experiment design, occurrence and
extent of distresses, and analyses methods used. A “weak” effect at this point in time may
become a “medium” or “strong” effect in the long term. Hence, all the conclusions are
based on “mid-term” performance of the ongoing SPS-1 experiment.

The synthesis of results is presented next for each performance measure

separately.

6. 8.1 Effects of structural factors for ﬂexible pavements — SPS-1 experiment

This section is subdivided into three parts: (i) pavement performance, (ii)
pavement response, and (iii) relationship between response and performance. The
structural factors include HMA thickness, base thickness, base type, and drainage. The
experiment also includes studying the secondary effects of site factors, namely subgrade

type and climatic zones.

6.8.1.1 Effect of Design and Site Factors on Pavement Performance

The effects of the experimental factors on each performance measure are discussed

below, one performance measure at a time.

Fatigue Cracking

All the experimental factors were found to be affecting fatigue cracking, though
not at the same level. On the whole, pavements with “thin” 4-inch (102 mm) HMA
surface layer have shown more fatigue cracking than those with “thick” 7-inch (178 mm)

HMA surface layer. Also pavements constructed with only dense-graded aggregate base

289

(DGAB) have shown more fatigue cracking than those with dense-graded asphalt treated
base over unbound aggregate base (ATB/DGAB) and those with ATB base only, with the
latter base type showing the best performance. The effects of HMA surface thickness and
base type were found to be statistically and practically signiﬁcant. The main effect of
base thickness was found to be statistically insigniﬁcant. However, on average,
pavements with l6-inch (406 mm) base thickness have shown slightly better fatigue
performance than those with 8-inch (203 mm) or 12—inch (305 mm) base thickness. It
should be noted that only pavement sections with drainage have a 16-inch (406 mm) base
thickness according to the SPS-1 experiment design; therefore, it is unclear whether this
effect is caused by the increased base thickness or by drainage provided with the
permeable asphalt treated base (PATB). In this regard, the frequency-based analyses did
show that pavements with drainage have signiﬁcantly lower chances of cracking than
those without drainage.

In general, pavement sections built on ﬁne—grained soils have more fatigue
cracking than those built on coarse-grained soils. Also pavements located WF zone have
shown more fatigue cracking than those located in WNF zone. These effects were found
to be statistically signiﬁcant and are of practical signiﬁcance.

Among un-drained pavements, on average, an increase in HMA surface thickness
from 4-inch (102 mm) to 7-inch (178 mm) has a slightly higher effect on fatigue cracking
for pavements with DGAB than for pavements with ATB.

The above effect of HMA surface thickness is more signiﬁcant for sections built
on coarse-grained soils. On the other hand, among pavements built on fme-grained soils,

the effect of drainage is seen only in those sections with DGAB; i.e., those with drainage

290

have less fatigue cracking than those without drainage. Also among drained pavements
built on ﬁne-grained soils, those with 8-inch (203 mm) base have more cracking than
those with 12-inch (305 mm) and 16-inch (406 mm) base. These effects were found to be
statistically and practically, signiﬁcant. Hence, for pavements built on ﬁne-grained soils,
drainage improves the fatigue performance, especially if built with thicker bases.

The main effect of HMA thickness, discussed above, is mainly seen among
sections located in WNF zone. The effect is of practical and statistical signiﬁcance. This
may be an indication that an increase of HMA thickness from 4-inch (102 mm) to 7-inch
(178 mm) is not sufﬁcient in resisting fatigue cracking for pavements in WF zone as
compared to WNF zone.

Among sections located in the WF zone, those with DGAB have shown the
highest amount of cracking while those with ATB have the least cracking. In addition, 1
those with 16-inch (406 mm) drained base have the least amount of fatigue cracking.
These effects were found to be statistically and practically signiﬁcant. This suggests that
among pavements located in WF zone, “thick” 16-inch (406 mm) treated bases with
drainage are less prone to cracking. The effects of HMA thickness and base thickness
discussed above imply that, among sections located in WF zone, an increase in base
thickness to 16-inch (with drainage) has a greater impact than an increase in HMA
thickness from 4-inch (102 mm) to 7-inch (178 mm), suggesting that a thicker base and

drainage helps in reducing frost effects.

291

Structural Rutting

The extent of structural rutting among the test sections in the SPS-1 experiment is
6.5 mm, on average, with a standard deviation of 2.4 mm. Their average age is about 7
years with a range between 4.5 and 10 years. The amount of rutting for the majority of
these sections is within the normal range at this point in time. Therefore, the results at this
point may only show initial trends and may not be of much practical signiﬁcance.

Marginal main effects of drainage, HMA thickness, and base thickness on
structural rutting were observed. Pavements with “thin” [4—inch (102 mm)] HMA surface
layer have shown slightly more rutting than those with “thick” [7—inch (178 mm)] HMA
surface layer. Also, on average, pavements with 16-inch (406 mm) drained base have
shown somewhat better rut performance than those with 8-inch (203 mm) and 12-inch
(305 mm) base. However, these effects of HMA surface thickness and base thickness
were not found to be statistically signiﬁcant. Pavements with drainage have less rutting
than those without drainage. The effect of drainage on structural rutting was found to be
statistically signiﬁcant; however the effect is not of practical signiﬁcance at this point in
time.

In general, pavement sections built on ﬁne-grained subgrade have shown more
rutting than those built on coarse-grained subgrade. This effect is statistically signiﬁcant
and appears to be of practical signiﬁcance. On the other hand, there is no apparent effect
of climate (W F vs. WNF) on structural rutting.

Among the pavements built on coarse-grained soils, those with 7—inch (178 mm)
HMA surface have shown slightly less rutting than those with 4-inch (102 mm) HMA

surface. This effect was statistically signiﬁcant; however it is not operationally

292

meaningful at this point. The above suggests that for sections built on ﬁne-grained soils
an increase in HMA thickness from 4-inch (102 mm) to 7-inch (178 mm) may not be
sufﬁcient in reducing the amount of rutting. On the other hand, among pavements built
on ﬁne-grained soils, a marginal positive effect of drainage is seen in sections with ATB.

Among drained pavements located in WF zone, those with DGAB have shown
more rutting than those with ATB. Also, among sections located in WF zone and built
with ATB, those with drainage have shown signiﬁcantly less rutting than those without
drainage. Both of these effects were found to be statistically signiﬁcant and are of
operational signiﬁcance. This implies that, among pavements located in WF zone, those
with ATB and drainage perform better than those with other combinations of base type
and drainage.

Among un-drained sections located in WNF zone, those with 12-inch (305 mm)
base have less rutting than those with 8-inch (203 mm) base. This effect was found to be
statistically signiﬁcant and of practically signiﬁcance. For sections built on DGAB and
located in WNF zone, those with drainage have shown slightly less rutting than those
without drainage. The effect was found to be marginally signiﬁcant. These early trends
imply that the importance of drainage among pavements with DGAB is considerable in
improving rut performance among sections located in WNF zone. On the other hand an
increase in base thickness from 8-inch (203 mm) to 12-inch (305 mm) improves rut

performance for un-drained sections, irrespective of base type.

Roughness
All the experimental factors were found to be affecting roughness, though not at

the same level. Pavements with “thin” [4-inch (102 mm)] HMA surface layer have higher

293

change in IRI (AIRI) than those with “thick” [7-inch (178 mm)] HMA surface layer. This
effect was found to be statistically signiﬁcant but is not of practical signiﬁcance at this
point in time. Also, pavements constructed with DGAB have higher AIRI than those with
ATB/DGAB and ATB, while pavements with ATB have the best performance for
roughness. Pavements with thicker bases have lower AIRI. Also pavements with drainage

have lower AIRI than un-drained pavements. The above main effects of base thickness,
base type and drainage were found to be statistically signiﬁcant and are of practical
signiﬁcance.

In general, pavements built on ﬁne-grained soils have shown higher AIRI than
those built on coarse-grained soils, especially among sections in WF zone. Also, the
change in roughness among sections located in WF zone is signiﬁcantly higher than' those
in WNF zone. These effects were found to be statistically signiﬁcant and are of practical
signiﬁcance.

Among pavements built on ﬁne-grained soils, an increase in HMA thickness from
4-inch (102 mm) to 7-inch (178 mm) has a signiﬁcant positive effect on change in
roughness. Also for un-drained pavements, those with ATB have signiﬁcantly lower AIRI
than those with DGAB. Finally the effect of drainage is signiﬁcant only for sections with
DGAB. The above effects were found to be statistically signiﬁcant and are of practical
signiﬁcance. These effects suggest that, for pavements built on ﬁne-grained soils, higher
HMA thickness and/or treated base will help inhibit the increase in roughness. Also,
drainage appears to be more effective in preventing an increase in roughness for sections

with DGAB, especially among those located in WF zone.

294

For un-drained pavements built on coarse-grained soils, an increase in base
thickness from 8-inch (203 mm) to 12-inch (305 mm) has a marginally signiﬁcant effect,
in that sections with thicker base have lower AIRI. However, this effect is not of practical

signiﬁcance at this point in time.

Transverse Cracking

The effect of base thickness on transverSe cracking is insigniﬁcant, at this point.
Pavements constructed with DGAB have more transverse cracking than those with
ATB/DGAB and ATB, while pavements with ATB have shown the least amount of
cracking. The effect was found to be statistically signiﬁcant; however it is not of practical
signiﬁcance at this point in time. Slightly more cracking was observed on pavements with
“thin” [4-inch (102 mm)] HMA surface layer. Also, pavements with drainage have shown
slightly less cracking than un-drained pavements. However, these effects were not found
to be statistically signiﬁcant.

In general, pavements built on ﬁne-grained soils have shown more transverse
cracking than those built on coarse-grained soils. This effect was found to be statistically
signiﬁcant and is of practical signiﬁcance.

Pavements located in WF zone have shown signiﬁcantly more transverse cracking
than those located in WNF zone. This main effect of climatic zone was found to be
statistically signiﬁcant and is of practical signiﬁcance. This conﬁrms that transverse
cracking occurs mainly in freezing environment.

Among drained pavements built on coarse-grained soils, those with ATB
performed better than those with DGAB. Also, among pavements with DGAB and built

on ﬁne-grained soils, those with drainage have shown signiﬁcantly less transverse

295

cracking than those without drainage. These effects were statistically signiﬁcant and

appear to be of practical signiﬁcance.

Longitudinal Cracking-WP

The effects of HMA and base thickness on longitudinal cracking—WP are
insigniﬁcant at this point in time. Pavements with drainage have shown less cracking than
un—drained pavements. The main effect of drainage was found to be statistically
signiﬁcant, but is not of practical signiﬁcance at this point.

In general, pavements built on ﬁne-grained soils have shown more longitudinal '
cracking-WP than those built on coarse-grained soils. This effect is of statistical and
practical signiﬁcance. Also, on average pavements in WF zone have shown higher levels
of longitudinal cracking-WP than those in WNF, especially among pavements built on
ﬁne-grained subgrade. This effect was found to be marginally signiﬁcant.

Among pavements built on ﬁne-grained soils, those built with DGAB have shown
more longitudinal cracking-WP, and those built with ATB have shown the least amount
of cracking. This main effect of base type was statistically and operationally signiﬁcant.
Also among pavements built on ﬁne-grained soils, drainage has a signiﬁcant effect on
longitudinal cracking, and this effect is more pronounced among pavements built with
DGAB. This effect was statistically signiﬁcant and is of practical signiﬁcance. This
trend implies that if a pavement on ﬁne-grained subgrade is constructed with a DGAB
base, better performance (in terms of longitudinal cracking-WP) can be achieved by

providing drainage. These effects are seen in both WF and. WNF zones.

296

Longitudinal Cracking-NWP

The effects of HMA thickness, base thickness, and base type on longitudinal
cracking-NWP are insigniﬁcant at this point in time. Pavements with drainage have
shown slightly less cracking than un-drained pavements. However, the effect of drainage
was found to be only marginally signiﬁcant.

The effect of subgrade type was not found to be statistically signiﬁcant. In
general, more longitudinal cracking-NWP was observed among sections located in
“freeze” climate compared to those in “no-freeze” climate. This main effect of climatic
zone is statistically signiﬁcant and is of practical signiﬁcance. Also, the effect of
drainage is more pronounced (with marginal statistical signiﬁcance) among pavements
located in “freeze” climate. However, this effect is not of practical signiﬁcance.

These initial trends indicate that longitudinal cracking-NWP is caused by “freeze”

climate (frost effects), and that pavements without drainage may be more prone to it.

6.8.1.2 Effect of Design and Site Factors on Pavement Response

Three pavement response parameters were chosen for AN OVA— peak deﬂection
under FWD load (do), far-sensor deﬂection (d6), and AREA. All the response parameters
have been calculated using the initial deﬂections of the test sections. Also, the pavement
surface temperature at the time of testing was taken as a covariate along with the age at
the time of testing and variability in the HMA and base layer thicknesses. The natural
logarithmic transformation has been applied to the three response indicators to fulﬁll the
ANOVA assumptions. The following discussion summarizes the effects of design and

site factors on each of the response parameters.

297

Peak Deﬂection under FWD Load (do)

The interactions between HMA thickness and base type, base thickness and base
type, base type and drainage, have signiﬁcant effects on the peak deﬂection (do).

Among the pavement sections built on DGAB, those with 4-inch (102 mm) HMA
thickness have higher'do than those with 7-inch (178 mm) HMA thickness. Also as
expected, thicker bases for each base type have lower do. However, this effect was more
signiﬁcant in the case of sections with treated bases (ATB or ATB/DGAB). Furthermore,
pavement sections with PATB/DGAB have lower do than those with DGAB.

The interaction between subgrade soil and climatic zone was found to have a very
signiﬁcant effect on do. Test sections built on ﬁne-grained soils have shown signiﬁcantly
higher do as compared to those built on coarse-grained soils. This effect is more

prominent on pavements located in WNF zone.

Far Sensor Deﬂection (d6)

The effects of base type, base thickness and drainage have signiﬁcant effects on the
far-sensor deﬂection (d6). HMA thickness and pavement mid depth temperature do not
have a signiﬁcant effect on d6.

The interaction between subgrade soil type and climatic zone was found to have a
signiﬁcant effect on (16. Test sections built on ﬁne-grained soils have shown signiﬁcantly
higher (16 as compared to those built on coarse-grained soils. This effect is more
prominent on pavements located in WNF zone.

Pavement sections built with DGAB have shown higher far-sensor deﬂections
than those built on other base types. Pavements constructed on 8-inch bases have also

shown signiﬁcantly higher far-sensor deﬂections than those built on 12-inch (203 mm) or

298

l6-inch (406 mm) bases. Furthermore, pavement sections with PATB/DGAB have lower
(16 than those with DGAB. These effects of the design factors on d6 are based on
statistical analyses only, and may or may not be of practical importance.

AREA

The interactions between HMA thickness and base type, base thickness and base
type, and, drainage and base type have signiﬁcant effects on the AREA parameter.

Among pavement sections built on DGAB, those with “thin” HMA surface layer
have lower AREA values compared to those with “thick” HMA surface layer, implying
that the upper layers of these pavement sections are “less stiff". The increase in HMA
thickness from 4-inch (102 mm) to 7-inch (178 mm) on ATB does not signiﬁcantly
increase the AREA value.

For sections built on DGAB, increasing base thickness from 8-inch (203 mm) to
12-inch (305 mm) has not shown a signiﬁcant effect on AREA; however a two-fold
increase in base thickness [from 8 tol6 inch (203 to 406 mm)] has shown‘a signiﬁcant
increase in AREA. Also, base thickness does not seem to have a signiﬁcant effect on
AREA in pavement sections with ATB bases. Furthermore, pavement sections with
PATB/DGAB have higher AREA values than those with DGAB. This indicates that the
structural capacity of the PATB layer is somewhat higher than that of the DGAB.

Among the pavement sections located in WNF zone, those built on ﬁne-grained
subgrade soils have signiﬁcantly higher AREA values than those built on coarse-grained
soils. However, in the case of sections located in WF zone, this effect is not signiﬁcant

indicating that AREA could be independent of the subgrade soil type.

299

A simpliﬁed summary of results from all analyses is given in Table 6-42. The summary is
only meant to give an overall assessment of the effects. The reader is strongly
recommended to read the following write-up for a better understanding of all the effects.
It is important to note that a “strong”, “medium” or “weak” effect should only be
interpreted in terms of the difference in effects at the various levels of a factor. As an
example, a “strong” effect of HMA thickness and a “strong” effect of subgrade soil type

should not be interpreted as HMA thickness and subgrade type having the same strength

of effect.

A black circle indicates a “strong” effect (signiﬁcant); a grey circle indicates a
“medium” effect, and a white circle indicates a “weak” effect. Operational signiﬁcance
was determined only for “strong” or “medium” effects. It should be noted that an effect
can be statistically signiﬁcant (meaning that it is not a coincidence) but may not be

operationally/ practically signiﬁcant, at this point in time.

300

 

PE #83

 

 

938 “out“. Ems: Hootm Macaw

Comte notoEBEV “ovum 83qu ®

 

:oumtomoo BoExm

 

 

 

 

.wﬁnzﬁﬂoucz .553
m .8 tone 2: E 38.. 2922 can: 8 new»: 2 532 2:. amaze .2959 a S. acute 05 mo 0E8 wENEEEa Lo omega 2: ._8 buxom a 28: £5 ”202

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

ab
. . . . . . ouMcwnsm
I I I I I I ammo
. a g . 9 G ummEEQ
I I o I e 9 ram...
. . . . Q . cab 3mm
.4 . 3.05.25

0 . 5% . G . <2:

o D m
:ocoomwn xaom cozoucwn xmom QMLHMMHWL. mmoccwsom mcEzm oﬂwwoamwu .56“;
3.5302 ammo mom 8:582 uocmctoﬁom canon

 

 

 

 

$58023 £25m “8 E98.“ 36 can. :wmmuw mo manta mo c5883 .uoEEEE. $6 035.

301

6.8.1.3 Apparent Relationship between Response and Performance

Two types of relations between ﬂexible pavement response under (FWD testing)
and performance were explored for the SPS-l pavement sections—explanatory and
predictive. Explanatory relationships were established using multiple regressions on data
from all the test sections in the experiment. Predictive relationships were established
based on bivariate correlation analyses at the site level, and using scatter plots on data
from all sections. The dynamic load response (DLR) data from instrumented sections in
Ohio were used for predictive relationships. The salient ﬁndings are brieﬂy presented

below:

Overall Analysis-—- Explanatory Relationship

A regression model was developed considering the peak deﬂection (do) as the
dependent variable and variables such as temperature, asphalt thickness, subgrade
strength and performance measures as independent variables. The observations based on
the regression model are as follows:

0 Pavements with “thick” [7-inch (178 mm)] HMA surface layer were observed (with
statistical signiﬁcance) to have signiﬁcantly lower deﬂections than those with “thin”
[4-inch (102 mm)] HMA surface layer.

0 Mid-depth temperature of the HMA layer, at the time of testing, has a statistically
signiﬁcant effect on do, Irrespective of design features, pavement deﬂections (do)
measured at higher temperatures is greater than those at lower temperatures.

0 Older pavements have slightly lower deﬂections (do) compared to younger

pavements, which could be due to stiffening (aging) of the asphalt.

302

o Pavements with “weaker” subgrade (higher do) have signiﬁcantly higher do (with
statistical signiﬁcance).

- Pavements with more cracking (fatigue cracking or longitudinal cracking) have a
signiﬁcantly higher do (with statistical signiﬁcance), compared to those with less
cracking.

Site Level Analysis— Predictive Relationships

This section summarizes the ﬁndings regarding the predictive relationships
between initial response (FWD deﬂection or deﬂection basin indices) and future
pavement performance (fatigue cracking, rutting and roughness) at the site level. The data
for sections from LA (22) were excludedfrom these analyses, as performance data for the
sections are available for just one year.

0 On average, AREA, SCI and BDI have shown reasonable correlations with
fatigue cracking for sections in most of the sites in the SPS-l experiment. In most
of the sites, pavements with higher initial SCI or BDI, or lower initial AREA were
found to have higher fatigue cracking.

o Consistent trends were observed between EDI and future IRI for the various sites
in the SPS-1 experiment. In most of the sites, pavements with higher initial BDI
were found to have higher IRI.

- The deﬂection basin parameters have not shown a consistent relationship with rut
depth for the various sites in the SPS-l experiment.

Overall Analysis— Predictive Relationships
Relationships were explored between initial response (FWD deﬂection basin

indices) and pavement performance (cracking, rutting and roughness), using bivariate

303

scatter plots between selected response parameters and performance measures for all

pavement sections in the experiment. The main observations based on these relationships

are listed below:

Among pavements constructed on fme-grained soils, ones with higher SCI
have shown more fatigue cracking, especially in WNF zone. Also, stiffer
pavements (higher AREA) on ﬁne-grained soils have shown more fatigue
cracking, especially if located in WF climatic zone.

Higher longitudinal cracking-WP was observed for'the pavement sections
with higher AREA especially among pavements located in WNF climatic
zone.

No apparent relation was observed between AREA and longitudinal cracking-
NWP, implying that this distress could be independent of the pavement
structural capacity.

No apparent trend was observed between AREA and transverse cracking.
This could imply that this distress type is not load-related.

Among pavements constructed on ﬁne-grained soils and located in WF zone,
those with higher BDI experienced slightly higher rutting. It was also
observed that some pavements with lower BDI (stronger structure) have
experienced higher rutting as compared to pavements with high BDI. These
pavements could have experienced mix-related rutting (not structural rutting).
Among pavements constructed on ﬁne-grained soils and located in WF zone,

those with higher BDI developed slightly higher roughness over time.

Dynamic Load Response for OH (39) test sections

This section of the report presents the summary of ﬁndings from the analysis of
measured Dynamic Load Response (DLR) data from the instrumented ﬂexible pavement
sections in the state of Ohio. The observations from the analysis of these instrumented
sections are summarized below:

0 In general, the strains in the longitudinal direction are higher than the strains in the
transverse direction; this is consistent with the results from mechanistic analysis of
ﬂexible pavements.

- The sections that were observed to have higher initial strain values have shown
worse fatigue performance. These results are in agreement with the mechanistic-
empirical design predictions that fatigue cracking in ﬂexible pavements is related to
the initial tensile strain at the bottom of the HMA layer (bottom up cracking).

o The sections that were observed to have high initial stress at the top of the subgrade
layer and those that were observed to have high initial surface deﬂection under the

load have shown poor rut performance.

305

CHAPTER 7 - ANALYSIS RESULTS FOR THE SPS-8
EXPERIMENT

7.1 INTRODUCTION

The purpose of this chapter is to provide a summary of the results of the analyses
conducted for the SPS-8 experiment on ﬂexible pavements. The performance measures
used in the analysis include fatigue cracking, rutting, longitudinal cracking (in the wheel
path and outside the wheel path), transverse cracking, and IRI. The results are
summarized according to individual design and site factors.

As mentioned in Chapter 4 under section 4.4, all the ﬂexible pavement sections in
SPS-8 experiment are aged between 3 and 10 years, with an average age of about 6 years.
Thus a majority of pavement sections are relatively “young” to exhibit any environment-
related distresses. Only a few sections have shown some distresses as of Release 17.0. It
is to be noted that the current status for SPS-8 experiment for ﬂexible pavements shows .
that there a few sites located in the DF and DNF zones. The extent of various distresses
exhibited by the pavements is presented in Chapter 4. Site-wise summaries of inventory
data, construction issues and performance of ﬂexible and rigid pavements can be found in
Appendix C. Keeping in view the number of sections constructed for SPS-8 experiment
(32 ﬂexible pavements in 15 sites) and the extent of distressesat present, statistical
analysis as in the case of SPS-1 experiment may not be applicable. Therefore, simple
mean comparisons (only for sections that exhibited distresses) were performed to identify
the effects of experimental factors on various performance measures. Some initial trends

obtained from these comparisons are reported below.

306

7.2 EFFECTS OF ENVIRONMENTAL FACTORS IN SPS-8 EXPERIMENT
FOR FLEXIBLE PAVEMENTS

The objective of the SPS-8 experiment is to develop conclusions concerning
environmentally induced serviceability loss and the contribution of environment and
subgrade to the distress of pavements. The experiment will also develop conclusions
concerning the effects of base and surface thickness variations on retarding

environmentally driven distress.

7.2.1 Site-Level Analysis

The analysis of the data from SPS—8 sections was done based on the concepts of

performance index (PI) and relative performance as in the case of SPS-l experiment. At

the site-level, various performance measures (fatigue cracking, longitudinal cracking
(WP and NWP), transverse cracking, raveling, rutting and roughness) were analyzed to
investigate the effects of the main site factors (climatic zone and subgrade type) on

performance. The summary of results from this analysis is given below:

0 The results of the available data indicate that WF zones have shown relatively higher

potential for fatigue cracking; however as expected, the magnitude of distress is not

signiﬁcant.
0 Results for longitudinal cracking-NWPwere inconclusive.
0 Transverse cracking occurred mainly in freeze zones.
0 There was higher amount of raveling observed in WNF zone.
0 Rutting performance was similar in all environments and for different subgrade

types; this is to be expected since rutting is essentially a load-related distress.

307

o The results of roughness in terms of IRI show that sections built in WF zones appear

to have higher roughness, followed by those built in WNF zones.

7.2.2 Overall Analysis

The overall initial trends which show the effect of SPS-8 experimental factors on
various performance measures will be discussed in this section. These comparisons were
carried out only for the performance measures which have shown some extent in the SPS-

8 ﬂexible pavements.

Fatigue Cracking: Fatigue cracking was observed in only 12 out of 32 pavement
sections. Among the cracked sections, the area of fatigue cracking varies from 0.2% to
about 19% with an average of 3%. Excluding section 36—0801, where 19% of area has
fatigue cracking, the average cracking area of cracked sections is about 1% with a range ,
0f 0.2% to 4.5%. The average fatigue cracking by experimental factors is shown in
Figure 7-1. The average fatigue performance shows that pavements in WF zone have a
higher potential for cracking. On average, pavements constructed on active subgrade
(frost susceptible or expansive) soils and pavements with “thin” [4-inch (102 mm)] HMA
surface layer have exhibited more cracking than those built on other subgrade types and
with “thick” [7-inch (178 mm)] HMA surface layer. In orderto show the effect of active -
subgrade within ﬁne and coarse soil types, the average performance is presented in
Figure 7-2. It was observed that ﬂexible pavements constructed on the active coarse-

grained subgrade soils have shown higher potential for fatigue cracking.

Longitudinal Cracking (WP and NWP): Longitudinal cracking-WP was observed in

only 13 out of 32 pavement sections, while longitudinal cracking-NWP occurred in 20

308

pavement sections. Among the cracked sections, longitudinal cracking-WP length varies
from 1 to 97 m with an average of 18 m, while longitudinal cracking-NWP length varies
from 1 to 305 m with an average of 115 m. Excluding both sections from MT (30) site,
where 78 m and 97 m of longitudinal cracking-WP occurred in sections 0805 and 0806,
among cracked sections, the average crack length is 5 m with a range of l to 22 m. The
average longitudinal cracking in the wheel path (WP) and non-wheel path (NWP) by
experimental factors are shown in Figure 7-3 and Figure 7-5. More longitudinal cracking-
WP is observed on sections located in DF zone. This cracking is mainly contributed by
sections in the MT (30) site, which are constructed on coarse subgrade type. More
longitudinal cracking-NWP is observed in all pavements constructed on active subgrade
soils and located in WF zone. Also more longitudinal cracking-NWP is observed in
sections located in DNF, which is contributed by sections in the NM (35) site only; these
sections were constructed on ﬁne-grained subgrade soils. The ﬂexible pavements
constructed on active ﬁne-grained soils have shown slightly more longitudinal cracking-
WP; the opposite trend is observed for pavements constructed on coarse-grained
subgrade; however this is due to the contribution of only one section at the MT (30) site
(see Figure 7-4). More longitudinal cracking-NWP is observed for ﬂexible pavements

constructed on active soils (see Figure 7-6).

Transverse Cracking: Transverse cracking was observed in only 10 out of 32 pavement
sections. Among the cracked sections, transverse cracking length varies from 1 to 44 m
with an average of 11 m. The average transverse cracking by experimental factors is
shown in Figure 7-7. Pavements with “thick” [7-inch (178 mm)] HMA surface layer have

shown less transverse cracking than those with “thin” [4-inch (102 mm)] HMA surface

309

layer. On average, more transverse cracking is exhibited by ﬂexible pavements
constructed on active soils and pavements located in “freeze” climates. The ﬂexible
pavements constructed on active subgrade have exhibited more transverse cracking than
those constructed on non-active subgrade soils especially within ﬁne subgrades (see

Figure 7-7).

Roughness: The average roughness, in terms of IRI, by experimental factors is presented
in Figure 7-9. The average initial IRI of the SPS-8 ﬂexible pavement sections is 1.1
m/km, with a range of 0.8 to 3.2 m/km. The average change in IRI (AIRI) for pavements
is 0.32 m/km with a range of 0.0 to 2.4 rn/km. Excluding both sections from OH (39) site,
where 2.2 m/km and 1.7 m/km of AIRI occurred in sections 0803 and 0804, the average
AIRI is 0.2 mfkm with a range of 0 to l m/km. On average, pavements located in “wet”
climate have higher change in IRI than those in “dry” climate. Furthermore, pavements
located in WF zone and those built on active soils have the higher changes in IRI. Also
pavements constructed on active subgrade soils have exhibited more roughness than other

pavements (see Figure 7-10).

Rut Depth: The average rut depth for ﬂexible pavements by experimental factors is
shown in Figure 7-11. The average latest rut depth of the sections is 5 mm with a range
of l to 24 mm. Excluding section 39-0803, where 24 mm of rut depth was observed, the
average rutting is about 4 mm with a range of 1 to 9 mm. On average, the magnitude of
rutting observed is “low” for the sections. Slightly higher average rut depths were

observed for the pavements located in WF zone.

310

35833 $58023 2“:ch

 

 

 

 

 

 

 

 

 

 

 

 

 

 

0::me @QO lamb owmcmnsm 3 $5.04 ommco>< v-5 oSmE @QO lﬂoﬁﬂ 23:08:35 .3 $5-04 omEo>< Tn 235
o>uo< I o>ﬂo<ucoz U E98,.“ Egﬁtoaxm
8C. oﬁansm 8:: =5 ..— o < “.20 ..E “23 ..E
U m o
ill _ O V
lulu t 2 v e .A.
W on m
1 on «a.
i ow m ow M
- 8 (wt 8 W
2. . i ll. E on
35833 £85m $5833 oimxoc m-mmm
wimm lamb ocacmnam B @5398 onwuﬂ owmco>< NH ocswmm ilmcouoﬁ Ecoﬁtoaxo .3 mat—once oswuﬁ owmb>< TN. Rama
333.. I o>=o<éoz D 298..“— _S=o§._onxm_
09¢. opacwnzm .
O m v55. £5. “— 0 < L75 mo ”—23 “—3
_ ll“- ,1. o v v
, 2 m a
- a m. m
on ...w
0
. ow m. w
.. om mw mm.
. 8 a a
E m. w

 

 

 

 

 

 

 

 

311

$56023 0338c $5533 29qu w- mmm

 

 

 

 

 

 

 

 

 

 

 

w- mmm IRS onﬁmpzm 3 was—080 8555.: owﬁo>< m. N. Saw—m [£98m 35:50on 3 was—8.8 88555 omSo>< \. n 8.an
0284‘. I o>_.o<..:oz U. Eoﬁﬁm Econ—toga
09C. avg—whom x25. 55. .mZD mn— LZB “—3
U m
[I 8
a N w
r m l v m
o m
- 2 W m m
3
u m— m 2 m
. N. x”
g z m
-i- I [Ea .-i? - R w I -- -1- i: - 2 N
358023 358ng 0::on ’

03:8: @QO [bub 3333. 3 E5264 omﬁo>< 0-x. Emmi w-mmm InmSBﬂ 32885on 3 952-04 owmuo>< 95 8:9;

 

 

 

 

 

 

 

959% I 03863.02 D 9.80am 35.53an
09:; oguwnzm x2e. SE. m “—20 un- ”—23 m3
0 .lIl-lrllLI 4 O
_ 0 on w
v m
T CW 8
m s. a.”
- om: .1
- on 0
T 02 m N
.OON M 8N m
I omm @ omm M
com I: }' l. in .Illii: com

 

312

$5833 35833 0:58:
2&on Wmmm Ilomb ocmhmnsm 3 Snow HE owﬁo>< N75 oBmE w-mmm .lmhogomm 358135 3 Snow a: ow80>< :-n 8:me
o>zo< I o>=o<éoZ D £28.”— BEoEEme
0&9 onaanzm .

 

 

 
 

 

 

 

 

 

 

 

 

 

 

U n— onP £5. “— 0 < L20 «.5 "~23 m3
. . ed
1 o _ .m am
1 ON 1 ON 9
. m 2 m
. . d
1 av ) 3 m.
I . 1 . w
I on m 3 m
r 0.0 o.o
$5833 $5838 0::on

0338.9 @QO lamb ocﬂwnsm 3 mmonsmse owﬁo>< 2-5 2ng $me Impoﬁﬂ 3581030 3 mmoéwse owﬁo>< 0S Bswi

v>zu< I u>co<rcoZ U 890mm 13:05.8me

ugh? ovwhwnzm
U m x85. 55. m O < mZQ “—0 m2? m3
o.o
r Nd
u v.0
I o.o
I wd
[ o;
I N;
Y. v.—
1 o._

c.
o

 

1
V?
C

o.—

 

 

unyur ‘I‘cII 982.19sz

(WW) 1211 ﬁes/xv

 
 

 

 

 

 

 

 

 

313

Also pavement sections with “thin” [4-inch (102 mm)] HMA surface layer have
exhibited higher rutting than pavements with “thick” [7-inch (178 mm)] HMA surface
layer. Furthermore, active subgrade seems to contribute towards more rutting (Figure
7-12). The magnitude of rutting was observed to be “low” for all sections in the

experiment, which is expected since rutting is essentially a load-related distress.

7.2.3 Comparison between Similar Designs of SPS-8 and SPS-l Experiments

To investigate the effect of traffic loading, similar designs in SPS-8 and SPS-l
experiments can be compared for various performance measures. This comparison may
help in identifying load-related and environment-related distresses.

The median comparisons (non-parametric) were performed on similar sections in
both experiments. From the experiment design matrix for SPS-1, sections 113 & 114 in
all states were identified to be similar to the structural design of SPS-8 ﬂexible pavement
sections. To investigate the median difference between the performances of these
sections, the sections were analyzed in two groups [4-inch (102 mm) and 7-inch (178
mm) HMA thickness].

It was found that fatigue cracking is essentially a load-related distress, while
transverse cracking and longitudinal cracking-NWP may be attributed to environment.

Table 7-1 shows the summary of results for these comparisons.

314

Table 7-1 Summary results for comparison between similar sections (SPS-8 vs. SPS—1)

 

 

 

 

 

 

 

 

 

 

Comparison P-value Remarks

There is statistical signiﬁcant difference
Alligator cracking for 4”HMA between median performances of sections
thickness 0.0058 within both experiments. This implies that
(SPS-8 vs. SPS- 1) alligator cracking is more critical to trafﬁc

loads.
Longitudinal cracking-WP for ghere is statciistical sirfgniﬁcant diffference
4”HM A thickness 0.0439 etzveen me ian pe ormagciles 0 sections
(SPS-8 vs. SPS-1) wt in both experiments. 15 shows that

this distress 15 load related.
Longitudinal cracking-NWP for ghere is no statistical signiﬁcant difference

, . etween median performances of sections

4 WA thickness 0.7068 . h' b th . t Th' th t
(SPS-8 vs. SPS-l) wt in o experimen s. 15 means a

this distress may not be load related.

There is no statistical signiﬁcant difference
Transverse cracking for 4”HMA between median performances of sections
thickness 0.7665 within both experiments. This means that
(SPS‘S vs. SPS-1) this distress may not be load related and can

be attributed to environment.

There is statistical signiﬁcant difference
Alligator cracking for 7” HMA between median performances of sections
thickness 0.0184 within both experiments. This implies that
(SPS-8 vs. SPS-l) alligator cracking is more critical to trafﬁc

loads.

There is no statistical signiﬁcant difference
Longitudinal cracking-WP for 7” between median performances of sections
HMA thickness 0.7022 within both experiments. This means that .
(SPS-8 vs. SPS-1) this distress may not be load related for

thicker pavement sections.

There is no statistical signiﬁcant difference
Longitudinal cracking-NWP for between median performances of sections
7” HMA thickness 0.1841 within both experiments. This means that
(SPS-8 vs. SPS-1) this distress may not be load related and can

be attributed to environment.
Transverse cracking for 7” HMA There is no statistical signiﬁcant difference
thickness N.A between median performances of sections

(SPS-8 vs. SPS-1)

 

 

within both experiments.

 

Note: All the comparisons were made at 0.05 level of signiﬁcance. N.A - Both median values are zero.

 

7.3 SUMMARY OF RESULTS

The SPS-8 experiment is entitled Strategic Study of Environmental Effects in
Absence of Heavy Loads for Flexible and Rigid Pavements. The study examines the
effect of climate and subgrade type (active, ﬁne, and coarse) on pavement sections
incorporating different ﬂexible and rigid pavements, which are subjected to very limited
trafﬁc as measured by ESAL accumulation. The effects of the experimental factors on

ﬂexible and rigid pavements, based on the initial trends, are summarized below.

7.3.1 Effect of SPS-8 Experimental Factors on Performance of Flexible Pavements

Currently a total of 32 ﬂexible pavement sections, in 15 sites, are present in the
experiment. There are 14, 8, 6 and 4 pavement sections in WF, WNF, DF and DNF
climatic zones, respectively. A total of 14 pavement sections were constructed on coarse-
grained soils among which 4 sections are on “active” soilsand 10 sections are on “non-
active” soils. Also, 18 pavement sections were built on fine-grained soils, among which
12 sections are on “active” soils and 6 sections are on “non-active” soils. These test
sections have an average age of about 6 years with a range of 3 to 10 years.

The SPS-8 pavements have “low” occurrence and extent of distresses, at this
point. Most of the pavements in the experiment are performing at comparable levels. No
formal statistical methods can be employed due to this. Therefore the observations
presented here are just based on average performance of the distressed pavements. The
observations, presented below, need to be considered as initial trends in light of these
limitations. The effects of the design and site factors based on initial trends, as of Release

17.0, on key performance measures are summarized below:

316

 

On average, pavements in WF zone have more fatigue cracking, longitudinal
cracking-NWP, and roughness than pavements in other climates. Also, in
general, pavements constructed on “active” subgrade (frost susceptible or
expansive) soils have higher longitudinal cracking-NWP, transverse cracking,
and fatigue cracking than pavements on “non-active” soils.

Pavements located in “wet” climate, on average, have higher change in IRI
than those in “dry” climate. Furthermore, pavements located in WF zone and

those built on active soils have the higher changes in IRI.

317

CHAPTER 8 - CONCLUSIONS AND RECOMMENDATIONS

8.1 INTRODUCTION

This chapter presents a summary of ﬁndings from a comprehensive evaluation of
SPS-land SPS-8 experiments, based on mid-term performance trends [Release 17 (Level-
B) of DataPave]. The current status of the experiment, construction quality, and data
availability are also brieﬂy discussed in this chapter for each experiment. Finally, the
limitations of the ﬁndings from this research and recommendations for future data
collection and research are presented.

A detailed description of the experiment designs and the current status of SPS-l,
and SPS-8 experiments were presented in Chapter 3. The SPS-1 experiment is a
fractional factorial experiment that was aimed at ﬁnding the relative inﬂuence of design
and construction features on performance of new ﬂexible pavements.

Each site within the SPS-1 experiment has twelve pavement test sections with
each section representing a different structural design. There are eighteen sites in the
SPS-1 experiment; these sites are distributed throughout the United States by climatic
zones (wet-freeze, wet-no-freeze, dry-freeze and dry-no-freeze) and subgrade type (ﬁne
and coarse-grained). The SPS-1 experiment is designed to investigate the effects of HMA
layer thickness, base type, base thickness, and drainage on ﬂexible pavement
performance. The current status (details in Chapter 3) of the experiments indicates some
deviations from the intended experiment design for SPS-1 experiment. The most
important deviation from the experiment design is the distribution of sites by climatic

zone, which caused an unbalance in the number of sites per climatic zone.

318

The average age of test sections in the SPS-1 experiment is 7 years with a range of 3 to
11 years. It may thus be said that the pavements are “fairly young”, and high occurrence
and levels of distresses may not be expected at this point in time. Thus, all conclusions
from the analyses presented in this report should only be interpreted as “mid-term”
performance ﬁndings.

The SPS-8 experiment was designed to study the effects of the environment on
pavement performance, in the absence of heavy trafﬁc. The experimental factors include
climate and subgrade soil type. A total of 32 ﬂexible pavement sections in 15 were
constructed for the experiment. The average age of the ﬂexible pavement sections in
SPS-8 is 6 years with a range of 3 to 10 years.

A summary of data availability was presented for SPS-land SPS-8 experiments in
Chapter 4. The extent and occurrence of distresses in the test sections within each of the
experiments were also presented in Chapter 4. Experimental factors (design features and
site factors) were compared to the as—constructed details obtained from the LTPP
database and from construction reports. The deviations observed were reported in the
chapter.

Based on the extent and occurrence of distresses, different methods of analysis
were employed. A brief description of each of the methods used for this research is in
Chapter 5. The majority of results from these analyses should be interpreted with caution
in light of the “low” occurrence of distresses in the test sections within the SPS-land
SPS-8 experiments. This is especially true for the SPS-8 experiment. Also, it is suggested
that the conclusions be considered while keeping in View the other limitations of the data,

as explained later in this chapter.

319

A synopsis of the salient ﬁndings from all the analyses presented in previous chapters
(Chapters 6 and 7), for each experiment, is presented in this chapter. This summary is
intended to give the reader a brief overview of the effects of design and construction
features on pavement performance and response.

The ﬁndings are presented by each performance or response measure, and not by
design factor, as most of the experimental factors (design and site factors) are interacting

among each other.

8.2 EFFECTS OF STRUCTURAL FACTORS FOR FLEXIBLE PAVEMENTS
-— SPS-1 EXPERIMENT

The SPS-1 experiment, entitled Strategic Study of Structural F actors for Flexible
Pavements, is one of the nine special pavement studies in the LTPP program. The effects
of the experimental factors on fatigue cracking, structural rutting, roughness, transverse
cracking, and longitudinal cracking (WP and NWP) are discussed below.

It should be noted that the effects presented herein are statistically signiﬁcant and

of practical significance unless mentioned otherwise.

8. 2.1 Eﬂect of Design and Site Factors on Pavement Performance

The effects of the experimental factors on each performance measure are discussed
below, one performance measure at a time.
Fatigue Cracking
o All the experimental factors were found to be affecting fatigue cracking,
though not at the same level. On the whole, pavements with “thin” 4-inch
(102 mm) HMA surface layer have shown more fatigue cracking than those

with “thick” 7-inch (178 mm) HMA surface layer. Also pavements

320

 

+“,—w ‘__- . -

constructed with only DGAB have shown more fatigue cracking than those
with ATB-over-DGAB, and those with ATB base only, with the latter base
type showing the best performance. The main effect of base thickness is not
statistically signiﬁcant. However, on average, pavements with 16-inch(406
mm) base thickness have shown slightly better fatigue performance than those
with 8-inch (203 mm) or 12-inch (305 mm) base thickness. It should be noted
that only pavement sections with drainage have a 16-inch (406 mm) base
thickness according to the SPS-1 experiment design; therefore, it is unclear
whether this effect is caused by the increased base thickness or by drainage
provided with PATB. In this regard, the frequency-based analyses did show
that pavements with drainage have signiﬁcantly lower chances of cracking
than those without drainage.

In general, pavement sections built on ﬁne-grained soils have more fatigue
cracking than those built on coarse-grained soils. Also pavements located WF
zone have shown more fatigue cracking than those located in WNF zone.
Among un-drained pavements, on average, an increase in HMA surface
thickness from 4-inch (102 mm) to 7-inch (178 mm) has a slightly higher
effect on fatigue cracking for pavements with DGAB than for pavements with
ATB. However, this effect is not statistically signiﬁcant.

The main effect of HMA surface thickness is more signiﬁcant for sections

built on coarse-grained soils.

Among pavements built on ﬁne-grained soils, the effect of drainage is seen

only in those sections with DGAB; i.e., those with drainage have less fatigue

321

 

cracking than those without drainage. For drained pavements built on ﬁne-
grained soils, those with 8-inch (203 mm) base have more cracking than those
with 12-inch (305 mm) and 16-inch (406 mm) base. Hence, for pavements
built on fme-grained soils, drainage improves the fatigue performance,
especially if built with thicker bases.

The main effect of HMA thickness, discussed above, is mainly seen among
sections located in WNF zone. This may be an indication that an increase of
HMA thickness from 4-inch (102 mm) to 7-inch (17 8 mm) is not sufﬁcient in
resisting fatigue cracking for pavements in WF zone as compared to WNF
zone.

Among sections located in the WF zone, those with DGAB have shown the
highest amount of cracking while those with ATB have the least cracking. In
addition, those with 16-inch (406 mm) drained base have the least amount of
fatigue cracking. This suggests that among pavements located in WF zone,
“thick” 16-inch (406 mm) treated bases with drainage are less prone to
cracking. The effects of HMA thickness and base thickness discussed above
imply that, among sections located in WF zone, an increase in base thickness
to l6-inch (with drainage) has a greater impact than an increase in HMA
thickness from 4-inch (102 mm) to 7-inch (178 mm), suggesting that a thicker

base and drainage helps in reducing frost effects.

322

Structural Rutting

The extent of structural rutting among the test sections in the SPS-l experiment is

6.5 mm, on average, with a standard deviation of 2.4 mm. Their average age is about 7

years with a range between 4.5 and 10 years. The amount of rutting for the majority of

these sections is within the normal range at this point in time. Therefore, the results at this

point may only show initial trends and may not be of much practical signiﬁcance.

Marginal main effects of drainage, HMA thickness, and base thickness on

structural rutting were observed. Pavements with “thin” [4-inch (102 mm)]
HMA surface layer have shown slightly more rutting than those with “thick”
[7-inch (178 mm)] HMA surface layer. Also, on average, pavements with 16-
inch (406 mm) drained base have shown somewhat better rut performance
than those with 8-inch (203 mm) and 12-inch (305 mm) base. However, these
effects of HMA surface thickness and base thickness were not found to be
statistically signiﬁcant. Pavements with drainage have less rutting than those
without drainage. The effect is not of practical signiﬁcance, at this point in
time.
In general, pavement sections built on ﬁne-grained subgrade have shown
more rutting than those built on coarse-grained subgrade. On the other hand,
there is no apparent effect of climate (WF vs. WNF) on structural rutting.
Among the pavements built on coarse-grained soils, those with 7-inch (178

mm) HMA surface have shown slightly less rutting than those with 4-inch

(102 mm) HMA surface. However, this effect is not operationally signiﬁcant

at this point.

323

 

o The above suggests that for sections built on ﬁne—grained soils an increase in
HMA thickness from 4—inch (102 mm) to 7-inch (178 mm) may not be
sufﬁcient in reducing the amount of rutting. Among pavements built on ﬁne-
grained soils, a marginal positive effect of drainage is seen in sections with
ATB.

0 Among drained pavements located in WP zone, those with DGAB have
shown more rutting than those with ATB. Also, among sections located in WF
zone and built with ATB, those with drainage have shown signiﬁcantly less
rutting than those without drainage. This implies that, among pavements
located in WF zone, those with ATB and drainage perform better than those
with other combinations of base type and drainage.

0 Among un-drained sections located in WNF zone, those with 12-inch (305
mm) base have less rutting than those with 8-inch (203 mm) base. For sections
built on DGAB and located in WNF zone, those with drainage have shown
slightly less rutting than those without drainage. This effect was found to be
marginally signiﬁcant. These early trends imply that the importance of
drainage among pavements with DGAB is considerable in improving rut
performance among sections located in WNF zone. On the other hand an
increase in base thickness from 8-inch (203 mm) to 12-inch (305 mm)
improves rut performance for un-drained sections, irrespective of base type.

Roughness ,
o All the experimental factors were found to be affecting roughness, though not

at the same level. Pavements with “thin” [4—inch (102 mm)] HMA surface

324

 

layer have higher change in IRI (AIRI) than those with “thick” [7-inch (178
mm)] HMA surface layer. This effect is not of practical signiﬁcance at this
point in time. Also, pavements constructed with DGAB have higher AIRI than
those with ATB/DGAB and ATB, while pavements with ATB have the best
performance for roughness. Pavements with thicker bases have lower AIRI.
Also pavements with drainage have lower AIRI than un-drained pavements.
In general, pavements built on ﬁne-grained soils have shown higher AIRI than
those built on coarse-grained soils, especially among sections in WF zone.
Also, the change in roughness among sections located in WF zone is
signiﬁcantly higher than those in WNF zone.

Among pavements built on ﬁne-grained soils, an increase in HMA thickness
from 4-inch (102 mm) to 7 -inch (178 mm) has a signiﬁcant positive effect on
change in roughness. Also for un-drained pavements, those with ATB have
signiﬁcantly lower AIRI than those with DGAB. Finally the effect of drainage
is signiﬁcant only for sections with DGAB. These effects suggest that, for
pavements built on ﬁne-grained soils, higher HMA thickness and/or treated
base will help inhibit the increase in roughness. Also, drainage appears to be
more effective in preventing an increase in roughness for sections with
DGAB, especially among those located in WF zone.

For un-drained pavements built on coarse-grained soils, an increase in base
thickness from 8-inch (203 mm) to 12-inch (305 mm) causes lower AIRI.
However, this effect is marginally signiﬁcant and is not of practical

signiﬁcance at this point in time.

325

Transverse Cracking

- The effect of base thickness on transverse cracking is insigniﬁcant, at this
point. Pavements constructed with DGAB have more transverse cracking than
those with ATB/DGAB and ATB, while pavements with ATB have shown the
least amount of cracking. However the effect is not of practical signiﬁcance at
this point in time. Slightly more cracking was observed on pavements with
“thin” [4-inch (102 mm)] HMA surface layer. Also, pavements with drainage
have shown slightly less cracking than un-drained pavements. However, these
effects were not found to be statistically signiﬁcant.

0 In general, pavements built on ﬁne—grained soils have shown more transverse
cracking than those built on coarse-grained soils.

- Pavements located in WF zone have shown signiﬁcantly more transverse
cracking than those located in WNF zone. This conﬁrms that transverse
cracking occurs mainly in freezing environment.

0 Among drained pavements built on coarse-grained soils, those with ATB
performed better than those with DGAB.

- Among pavements with DGAB and built on fine-grained soils, those with
drainage have shown signiﬁcantly less transverse cracking than those without
drainage.

Longitudinal Cracking-WP
o The effects of HMA and base thickness on longitudinal cracking-WP are

insigniﬁcant at this point in time. Pavements with drainage have shown less

326

 

cracking than un-drained pavements. The main effect of drainage is not of
practical signiﬁcance at this point.

In general, pavements built on ﬁne-grained soils have shown more
longitudinal cracking-WP than those built on coarse-grained soils.

On average pavements in WF zone have shown higher levels of longitudinal
cracking-WP than those in WNF, especially among pavements built on fme-
grained subgrade. This effect was found to be only marginally signiﬁcant.
Among pavements built on ﬁne-grained soils, those built with DGAB have
shown more longitudinal cracking-WP, and those built with ATB have shown
the least amount of cracking. Also, drainage has a signiﬁcant effect on
longitudinal cracking, and this effect is more pronounced in pavements built
with DGAB. This trend implies that if a pavement on ﬁne-grained subgrade is
constructed with a DGAB base, better performance (in terms of longitudinal

cracking-WP) can be achieved by providing drainage. These effects are seen

in both WP and. WNF zones.

Longitudinal Cracking-NWP

The effects of HMA thickness, base thickness, and base type on longitudinal
cracking-NWP are insigniﬁcant at this point in time. Pavements with drainage
have shown slightly less cracking than un-drained pavements. However, the
effect of drainage was found to be only marginally signiﬁcant.

The effect of subgrade type was not found to be statistically signiﬁcant.

In general, more longitudinal cracking-NWP was observed among sections

located in “freeze” climate compared to those in “no-freeze” climate.

327

o The effect of drainage is more pronounced (with marginal statistical
signiﬁcance) among pavements located in “freeze” climate. However, this
effect is not of practical signiﬁcance.

These initial trends indicate that longitudinal cracking-NWP is caused by “freeze”

climate (frost effects), and that pavements without drainage may be more prone to it.

In summary, based on the above discussion for SPS-l experiment, base type
seems to be the most critical design factor for fatigue cracking, roughness (IRI), and
longitudinal cracking-WP. This is not to say that the effect of HMA surface thickness is
not signiﬁcant. In fact, the effect of base type should be interpreted in light of the fact
that a dense graded asphalt treated base effectively means thicker HMA layer. Drainage
and base type, when combined also play an important role in improving ﬂexible
pavement performance, especially in terms of fatigue and longitudinal cracking. Base
thickness has secondary effects on performance, especially in the case of roughness and
rutting.

Subgrade soil type seems to be playing an important role in ﬂexible pavement
performance. In general, pavements built on ﬁne-grained soils have shown worst
performance, especially in the case of roughness. Also, climate is a critical factor in
determining ﬂexible pavement performance. Longitudinal cracking-NWP, transverse
cracking, and longitudinal cracking-WP appear to be affected by climate. Longitudinal
cracking-WP and transverse cracking seems to be associated with Wet Freeze
environment, while longitudinal cracking-NWP seems to be the dominant in “freeze”

climate.

328

8. 2.2 Effect of Design and Site Factors on Pavement Response

Three pavement response parameters were chosen for AN OVA - peak deﬂection
under FWD load (do), far-sensor deﬂection (do), and AREA. A summary of the effects of
design and site factors on each of the response parameters follows.

Peak Deﬂection under FWD Load (do): For pavement sections built on DGAB, those
with 4-inch (102 mm) HMA surface thickness have higher do than those with 7-inch (178
mm) HMA surface thickness. Also pavements with thicker bases, irrespective of base
type, have lower do. This effect is more prominent in the case of sections with treated
bases (ATB or ATB/DGAB). Also, pavement sections with PATB/DGAB have lower do
than those with DGAB.

In general, pavements built on ﬁne-grained soils have shown signiﬁcantly higher do
as compared to those built on coarse-grained soils. This effect is more prominent on
pavements located in WNF zone.

Far Sensor Deflection (d6): Pavements built on ﬁne-grained soils have higher do values
as compared to those built on coarse-grained soils. This effect is more prominent on
pavements located in WNF zone.

Pavements built with DGAB have shown higher do values than those builton other
base types. Pavements constructed on 8-inch (203 mm) bases have also shown
signiﬁcantly higher do values than those built on 12-inch (305 mm) or 16-inch (406 mm)
bases. Furthermore, pavements with PATB/DGAB have smaller d5 values than those with
DGAB. These effects of the design factors on do are based on statistical analyses only,

and may or may not be of practical importance.

329

AREA: For pavements built on DGAB, those with “thin” HMA surface layer have lower
AREA values compared to those with “thick” HMA surface layer, implying that the
upper layers of these pavements are “less stiff”.

For pavements built on DGAB, increasing base thickness from 8-inch (203 mm) to
12-inch (305 mm) has not shown a signiﬁcant effect on AREA; however a two-fold
increase in base thickness [from 8 tol6 inch (203 to 406 mm)] has shown a signiﬁcant
increase in AREA. Furthermore, pavements with PATB/DGAB have higher AREA
values than those with DGAB. This may be an indication that the structural capacity of

the PATB layer is somewhat higher than that of the DGAB.

8.2.3 Apparent Relationship between Response and Performance

Two types of relations between ﬂexible pavement response (FWD) and
performance were explored for the SPS-l pavements— explanatory and predictive. The

salient ﬁndings are brieﬂy presented below:

Overall Analysis— Explanatory Relationship
Following are ﬁndings regarding relationship between response and performance
based on regression analysis, after adjusting for HMA surface thickness and pavement
mid-depth temperature at the time of testing:
0 Older pavements have slightly lower deﬂections (do) compared to younger
pavements, which could be due to stiffening (aging) of the asphalt.
0 Pavements with “weaker” subgrade (higher do) have higher do values.
0 Pavements with more cracking (fatigue cracking or longitudinal cracking) have a

higher do values, compared to those with less cracking.

330

 

Site Level Analysis— Predictive Relationships

This section summarizes the observations regarding the predictive relationships

between initial response and future pavement performance, based on site-level analysis.

In most of the sites, pavements with higher initial SCI or BDI, or lower initial
AREA have higher fatigue cracking.

In most of the sites, pavements with higher initial BDI have higher IRI.

The deﬂection basin parameters have not shown a consistent relationship with rut

depth for the various sites in the SPS-1 experiment.

Overall Analysis— Predictive Relationships

The main observations based on the analyses are as follows:

0 For pavements constructed on ﬁne-grained soils, ones With higher SCI have

shown more fatigue cracking, especially in WNF zone.

Stiffer pavements (higher AREA) built on fme-grained soils have shown more
fatigue cracking, especially if located in WF climatic zone.

Higher longitudinal cracking-WP was observed for the pavement sections with
higher AREA, especially among pavements located in WNF climatic zone.

No apparent relation was observed between AREA and longitudinal cracking-
NWP or transverse cracking, implying that these distresses could be independent

of the pavement structural capacity.

Dynamic Load Response for OH (39) test sections

The observations based on analysis of DLR data from instrumented sections are:

In general, the strains in the longitudinal direction'are higher than the strains in

the transverse direction.

331

 

o The sections that were observed to have higher initial strain values have shown
worse fatigue performance.

0 The sections that were observed to have high initial stress at the top of the
subgrade layer and those that were observed to have high initial surface deﬂection

under the load have shown poor rut performance.

8.3 EFFECTS OF THE ENVIRONMENT IN THE ABSENCE OF HEAVY
TRAFFIC FOR FLEXIBLE PAVEMENTS — SPS-8 EXPERIMENT

The SPS-8 experiment is entitled Strategic Study of Environmental Eﬂects in the
Absence of Heavy Loads for Flexible and Rigid Pavements. The study examines the
effect of climate and subgrade type (active, ﬁne, and coarse) on pavement sections
incorporating different ﬂexible and rigid pavements, which are subjected to very limited
trafﬁc as measured by ESAL accumulation.

The SPS-8 pavements have “low” occurrence and extent of distresses, at this
point. Most of the pavements in the experiment are performing at comparable levels. No
formal statistical methods can be employed due to this. Therefore the observations
presented here are just based on average performance of the distressed pavements. The
observations, presented below, need to be considered as initial trends in light of these
limitations.

On average, pavements in WF zone have more fatigue cracking, longitudinal
cracking-NWP, and roughness than pavements in other climates. Also, in general,
pavements constructed on “active” subgrade (frost susceptible or expansive) soils have
higher longitudinal cracking-NWP, transverse cracking, and fatigue cracking than

pavements on “non-active” soils.

332

 

 

Pavements located in “wet” climate, on average, have higher change in IRI than those in

“dry” climate. Furthermore, pavements located in WF zone and those built on active soils

have the higher changes in IRI.

8.4 LIMITATIONS OF THE EXPERIMENTS AND ANALYSES

All the above ﬁndings/observations on the effects of design and construction

features on pavement performance and response should be considered in light of the

limitations discussed herein. These limitations can be broadly classiﬁed under two

categories— experiment-related and data-related.

8. 4. 1 Experiment-related issues

The SPS-land SPS-8 experiments, which are fractional-factorial designs, were
rendered unbalanced because unequal numbers of sites were constructed in each
zone-subgrade combination. This unbalanced design limits the “power” of the
experiments. In the SPS-1 experiment only 2 sites each are located in DF and
DNF zones, compared to 8 and 6 sites in WP and WNF zones. Moreover, both the
sites of SPS-l in the DF zone are located on coarse-grained soils. Furthermore, in
some of the sites not all sections were constructed on the same subgrade soil type
[for example, KS (20)].

Initially, the SPS-1 experiment was designed to have all 24 designs at a site. But
later due to some implementation issues, 12 designs were constructed per site.
Hence, the experiments do not have any “true” (statistical) replication of designs.
In other words, though two sites (with 12 designs at each site) are located in a
climate-subgrade soil combination, the trafﬁc, age, and material-related properties

vary between the sites.
333

8.4.2

The variation in age of the sites is considerably high for the experiments. If the
sites were reasonably similar in age, the ﬁndings would be more reliable. It
should be noted that age of the test sections was included as a covariate in all
statistical analyses to address the above issue to some extent.

In the SPS-1 experiment, in-pavement drainage was provided only for sections
built with PATB. Moreover, all sections with PATB were provided with drainage.
As a result of this, the effect of PATB and the effect of drainage are inseparable
(confounded).

In the SPS-1 experiment, a 16-inch (406 mm) thick base was only provided for
sections with drainage. In other words, none of the sections without drainage have
a 16-inch (406 mm) base thickness. Hence, the effect of a 16-inch (406 mm) thick
base and the effect of drainage are also inseparable.

Among ﬂexible pavement sections of SPS-l, all sections built with ATB-over-
DGAB were not provided with drainage. Hence, any interaction effects of
drainage and ATB—over-DGAB cannot be studied.

Only the lower limit for trafﬁc volume was speciﬁed for the SPS-l experiment.
This resulted in considerable variability in trafﬁc across sites.

In the SPS-8 experiment, the effects of HMA surface thickness and base thickness
are confounded. Therefore, the “pure” effects of any of these factors cannot be

studied.

Data-related issues

Reasonably accurate monitored trafﬁc data is not available for all the sites in the

experiments. This has further complicated the issue of controlling for trafﬁc.

334

 

0 Large measurement variability was observed, over time, for some of the distresses
(for example, longitudinal cracking in SPS-l). This variability has made the time-
series trends unclear for some of the performance measures.

0 The measurement variability discussed above is believed to be due to
maintenance activity at some sites, which is a deviation from the experiment
design. These activities tamper with the actual long-term ﬁeld performance of the
pavement sections.

0 The frequency of distress surveys is not uniform across sites, especially for SPS-l
experiments. Wide gaps in distress surveys necessitate interpolation of
performance, which may not always be accurate.

0 Though thorough construction guidelines were developed by the LTPP to
minimize construction variability across sites, some deviations occurred. These
deviations along with the material variability have added to the variability in
performance across sites. If material-related information were available for all the
sections, the issues caused by performance variability could be better addressed.

- Backcalculated layer moduli are unavailable for most of the sections in the SPS-1
experiment. Some of the material-related issues could be dealt with if the data

were present.

8.5 RECOMMENDATIONS FOR FUTURE DATA COLLECTION AND
RESEARCH

Based on the above issues and the experience with the LTPP data,

recommendations for future data collection and research are given below.

335

l)

2)

3)

4)

5)

6)

7)

Reasonably accurate monitored trafﬁc data should be made available for all the
sites to allow for better adjustment of trafﬁc loading variation across sites.

All the core sections of the SPS-1 experiment should be closely monitored until
failure or to a stage when the long-term performance (at least 15-20 years) has
been captured.

The core sections of the SPS-lexperiment should be strictly supervised to prevent
any maintenance activity, as per the experiment designs. This will ensure that the
actual long-term performance of the pavements is observed.

Most of the test sections will soon enter a critical stage in their service life; in
light of this, to reap maximum beneﬁts from the experiments, the sections should
be monitored at regular intervals and with greater accuracy.

Hall et a1 [1] have identiﬁed issues regarding in-pavement drainage for sections in
SPS-1 and SPS-2 experiments. The ﬁndings from this study should be considered
for inclusion in the DataPave IMS database. This may help study the effect of in-
pavement drainage more accurately.

Some of the sections in the SPS-l experiment have shown premature “failure”.
These sections should be considered for exclusion from DataPave, as they do not
contribute to the study of long-term pavement performance.

Some of the sites in the SPS-l experiment are very close to the thresholds
(regarding average annual precipitation and freeze index) deﬁned for delineation
of climatic zones. The deﬁnitions of the climatic zones may need reconsideration

in light of this.

336

 

8) Accurate material data should be made available for all the sections of the
experiments to allow for addressing the variability in material quality across sites.
Also, backcalculated layer moduli data should be made available for all sections
in DataPave to help perform mechanistic analyses.

9) Most of the construction-related issues are available only from construction
reports. These issues and/or deviations should be better highlighted well within
the DataPave database.

10) The spatial location of some distresses (such as cracking) is sometimes important
for research. It is practically cumbersome for the users to obtain distress maps and
interpret the spatial location of the distresses. It is therefore recommended that
each section be “discretized” (segmented) for data collection and the related data
be made available in DataPave. This would greatly decrease the level of
subjectivity in the data.

11) In general, the extent of distresses on the SPS-8 test sections is “low”, at this point
in time. The performance data should thus be collected for sufﬁciently long time
(15 to 20 years for all the sections in the experiment) to capture the effects of
environment. A meaningful statistical analysis may then be performed to study
the effects of environment.

12) The DLR instrumentation location (spatial location), alignment and designation
details should be made more accurate in DataPave.

13) It is recommended that the complete actual traces of data from the instrumented
DLR sections be considered for inclusion in the DataPave database (in addition to

the “peak” and “valley” data) after proper quality checks. Also more data (i.e.

337

 

more test series results) should be included in the DataPave. This data could be

stored separately as in the case of proﬁles and FWD time histories.

When the long-term performance data is available for most of the sections of the
SPS-l and SPS-8 experiments, the methods employed in this research would be more
“powerful” to study the effect of design and construction features. Methods that analyze
the time-series data (such as survival analysis and AN OVA with repeated measures) can
also be employed when the performance data for most of the sections is available for

about 15 years.

338

 

REFERENCES

CHAPTER 1

1. National Research Council, S., Speciﬁc Pavement Studies Experimental Design
and Research Plan for Experiment SPS-l, Strategic Study of Structural Factors for
Flexible Pavements. Feb, 1990, SHRP: Washington, DC.

CHAPTER 2

l. Embacher, R.A., M.B. Snyder, and TD. Odden, Using the Minnesota accelerated
loading facility to test retroﬁt dowel load transfer systems. Transportation
Research Record, 2001(1769): p. 134-141.

2. Haung, Y.H., Pavement Analysis and Design. 2nd Edition ed. 2004, Upper Saddle
River, NJ 07458: Pearson Prentice Hall.

3. Yoder, El. and M.W. Witczak, Principles of Pavement Design. 1975, John Wiley
& Sons, Inc.

4. Dempsey, B.J., Laboratory and Field Studies of Channeling and Pumping.
Transportation Research Record. 849.

5. Quintus, H.V. and AL. Simpson, Structural Factors for Flexible Pavements-
Initial Evaluation of the SPS-l Experiment, Final Report. 2001, FHWA:
Washington, DC.

6. Hall, T.K., Effectiveness of Subsurface Drainage for HMA and PCC Pavements.
2002, National Corporative Highway Research Program: Washington, DC.

7. Hall, T.K. and BC. Carlos, Effects of Subsurface Drainage on Performance of
Asphalt and Concrete Pavements. 2003.

8. Mladenovic, G., Y.J. Jiang, and MI. Darter, Study of Environmental Effects in
the Absence of the Heavy Trafﬁc Load-SPS-8. 2002, FHW A: Washington, DC.

9. Perera, R.W. and SD. Kohn, LTPP Data Analysis: Factors Affecting Pavement
Smoothness. 2001, BRE.

10. Rauhut, J .B., A. Eltahan, and AL. Simpson, Common Characteristics of Good
and Poorly Performing AC Pavements. 1999, FHWA.

11. Perera, R.W., C. Byrum, and SD. Kohn, Investigation of Development of
Pavement Roughness. 1998, FHWA, FHWA-RD-97-l47: Washington, DC.

12. Ali, AH. and D.T. Shiraz, Mechanistic Evaluation of Test Data from LTPP

Flexible Pavement Test Sections. 1998.

339

13.

14.

15.

16.
17.
18.

19.

Quintus, H.V. and K. Brian, Analysis Relating to Pavement Material
Characterizations and their Effects on Pavement Performance. 1998, FHWA:
Washington, DC.

Kim, R.Y. and SR. Ranjithan, Assessing Pavement Layer Condition Using
Deﬂection Data. 2000, NCHRP 10-48.

Stubstad, R.N., et al., LTPP Data Analysis: Feasibility of Using FWD Deﬂection
Data to Characterize Pavement Construction Quality. 2002, NCHRP, NCHRP's
Project 20-50(9), Final Report.

AASHTO, Guide for Design of Pavement Structures. 1993, American Association
of State Highway and Transportation Ofﬁcials: Washington, DC.

Ullidtz, P., Pavement Analysis. 1987, Amsterdam, The Netherlands: Elsevier
Science Publisher B.V.

Ullidtz, P., Modeling Flexible Pavement Response and Performance. 1998,
Gylling, Denmark: Narayana Press.

Simpson, AL, et al., Sensitivity Analysis for Selected Pavement Distresses.
1994, Strategic Highway Research Program, National Research Council:
Washington, DC.

CHAPTER 3

1.

Hanna, A.N., S.D. Tayabji., and J .S. Miller, SHRP-LTPP Speciﬁc Pavement
Studies: Five-Year Report. 1994, Strategic Highway Research Program, National
Research Council. ' ' '

National Research Council, S., Speciﬁc Pavement Studies Experimental Design
and Research Plan for Experiment SPS-1, Strategic Study of Structural Factors for
Flexible Pavements. Feb, 1990, SHRP: Washington, DC.

Kutner, M.H., et al., Applied Linear Models. 5th ed. 2005: McGraw-Hill, Irwin.

Mee, R.W., Efﬁcient Two-Level Designs for Estimating All Main Effects and
Two-Factor Interactions. Journal of Quality Technology, 2004. 36(No. 4): p. 400-
412.

Chatti, K., et al., Final Report NCHRP 20-50 (10/16), LTPP Data Analysis:
Inﬂuence of Design and Construction Features on the Response and Performance

of New Flexible and Rigid Pavements. 2005, Michigan State University: East
Lansing.

Quintus, H.V. and AL. Simpson, Structural Factors for Flexible Pavements-
Initial Evaluation of the SPS-1 Experiment, Final Report. 2001, FHWA:
Washington, DC.

340

 

CHAPTER 4

1. Chatti, K., et al., Final Report NCHRP 20-50 (10/ 16), LTPP Data Analysis:
Inﬂuence of Design and Construction Features on the ReSponse and Performance
of New Flexible and Rigid Pavements. 2005, Michigan State University: East
Lansing.

2. Quintus, H.V. and AL. Simpson, Structural Factors for Flexible Pavements-
Initial Evaluation of the SPS-1 Experiment, Final Report. 2001, FHWA:
Washington, DC.

3. Hall, T.K. and EC. Carlos, Effects of Subsurface Drainage on Performance of
Asphalt and Concrete Pavements. 2003.

4. Chen, D.H., et al., Forensic Evaluation of Premature Failures of Texas Speciﬁc
Pavement Study—l Sections. Journal of Performance and Constructed Facilities,
2003. Vol. 17(No. 2): p. 67-74.

5. Hanna, A.N., S.D. Tayabji., and J .S. Miller, SHRP-LTPP Speciﬁc Pavement
Studies: Five-Year Report. 1994, Strategic Highway Research Program, National
Research Council. .

6. ORITE, Continued Monitoring of Instrumented Pavement in Ohio. 2001, Institute -
for Transportation and the Environment (ORITE), Ohio University: Ohio.

7. Holtz, RD. and W.D. Kovacs, An Introduction to Geotechnical Engineering.
1981 : Prentice-Hall.

CHPATER 5

1. Rauhut, J .B., A. Eltahan, and AL. Simpson, Common Characteristics of Good
and Poorly PerfOrming AC Pavements. 1999, FHWA.

2. Khazanovich, L., et al., Common Characteristics of Good and Poorly Performing
PCC Pavements. 1998, FHWA.

3. Quintus, H.V. and AL. Simpson, Structural Factors for Flexible Pavements-
Initial Evaluation of the SPS-l Experiment, Final Report. 2001, FHWA:
Washington, DC. ‘

4. Hall, T.K. and EC. Carlos, Effects of Subsurface Drainage on Performance of
Asphalt and Concrete Pavements. 2003.

S. Larsen, J .R. and L.M. Morris, An Introduction to Mathematical Statistics and Its
Applications. 3rd ed. 2001: Prentice Hall.

6. Kutner, M.H., et al., Applied Linear Models. 5th ed. 2005: McGraw-Hill, Irwin.

7. Dillon, W.R. and G. Matthew, Multivariate Analysis-Methods and Applications.

1984, New York: John Wiley & Sons, Inc.

341

 

8. Ott, R.L. and L. Michael, An Introduction to Statistical Methods and Data
Analysis. 5th Edition ed. 2001, CA: Duxbury.

9. Khazanovich, L., et al., Common Characteristics of Good and Poorly Performing
PCC Pavements. 1998, FHWA.

10. Stroup, W.W., Mixed Models Procedures to Assess Power, Precision and Sample
Size in the Design of Experiments. 2004: University of Nebraska, Lincoln, NE.

11. Neter, J., et al., Applied Linear Statistical Models. 4th ed. 1996, Illinois: Richard
D. Irwin, Inc.

12. Hastie, T., T. Tobert, and J. Friedman, The Elements of Statistical Learning: Data
Mining, Inference, and Prediction. 2001, New York: Springer-Verlag.

13. Hair, J .F., Jr., et al., Multivariate Data Analysis. 5th ed. 1998, New Jersey:
Prentice Hall.

14. Hosmer, D.W. and L. Stanley, Applied Logistic Regression. 2nd ed. 2000, NY.:
John Wiley & Sons, Inc.

CHAPTER 6

1. Quintus, H.V. and AI. Simpson, Structural Factors for Flexible Pavements-
Initial Evaluation of the SPS-1 Experiment, Final Report. 2001, FHWA:
Washington, DC. ‘

2. Perera, R.W. and SD. Kohn, LTPP Data Analysis: Factors Affecting Pavement
Smoothness. 2001, BRE.

3. Hall, T.K., Effectiveness of Subsurface Drainage for HMA and PCC Pavements.
2002, National Corporative Highway Research Program: Washington, DC.

4. Hall, T.K. and EC. Carlos, Effects of Subsurface Drainage on Performance of
Asphalt and Concrete Pavements. 2003.

5. White, T.D., et al., Contribution of Pavement Structural Layers to Rutting of Hot
Mix Asphalt Pavements. 2002, Purdue University: West Lafayette, IN.

6. Chen, D.H., et al., Forensic Evaluation of Premature Failures of Texas Speciﬁc
Pavement Study-1 Sections. Journal of Performance and Constructed Facilities,
2003. Vol. 17(No. 2): p. 67-74.

7. Chatti, K., et al., Final Report NCHRP 20-50 (10/ 16), LTPP Data Analysis:

Inﬂuence of Design and Construction Features on the Response and Performance
of New Flexible and Rigid Pavements. 2005, Michigan State University: East
Lansing.

342

8. Stubstad, R.N., et al., LTPP Data Analysis: Feasibility of Using FWD Deﬂection
Data to Characterize Pavement Construction Quality. 2002, NCHRP, NCHRP'S
Project 20-50(9), Final Report.

9. Kim, R.Y. and SR. Ranjithan, Assessing Pavement Layer Condition Using
Deﬂection Data. 2000, NCHRP 10-48.

10. _ Lukanen, E.O., R. Stubstad, and R. Briggs, Temperature Predictions and
Adjustment Factors for Asphalt Pavements. 2000, FHWA, FHWA-RD-98-O85:
Washington, DC.

Chapter 8
1. Hall, T.K. and EC. Carlos, Effects of Subsurface Drainage on Performance of
Asphalt and Concrete Pavements. 2003.

343

 

lllllllllzllllllllllljljlll111111111