.....fimmm.
x . hr. ..

.. .332.

I . . I»
v 5.. . .
. wfir‘. .,. .. . . H.
c n .. :«Ibn r .... _ .. .
. u a. an. ..H . .. «v.5.
.. . .. 47...... . ...}... . . .
gum... .. _ . . 3...th .. aw v w. a. hwyuh...‘
' 'fia‘ . u I 3 0‘4 4.‘ v‘ ‘ ‘-.n.§-| 7
A‘ v13 . . .f W
m.

_ .... ... u u 143...}... a .. . . . . ...
. . . ... ...... . . ... . . 25...... ..,._...,M.....-...m.m.un.u1 . . _ . . e... ,.
.. ., .. . _ ...... _ . ... . ‘ .w . ...,mi. .. _ .
H . _ . a : . _. .. , a 4 aka... .... ....n. , , ......Jfiu...
, ... a. t. . .. . . ...: . . . . . ... .

....mhnfiisfl mafia
. .fiv. .u.. 91%. \
Emma?“ . . . o .. . .,
hit... 1.2.. .\ (.... .

V‘I
......fi

0 ..E .

. . . ...v . . . . . . ..
......«w... .29.... V . . . . .. . . . ; . .. . .... . a... .
. lanky. , . . . .. . . . . . . . .. : . . . .. . . L4Lv.-....l.mi«&...

Ea . . .. .. . . _ . . ......1.....H..r.:~.¢......n.n.......h....v

.
. . . . . L . . . . . . ... . . a . ... .-.
v .. . . . . .. .. . . . 2.9. ......u... . . . gnu”... Its»!!!
. .. , . . -.., ... .7 - . ... , . ... H. 2 -
.... . . .n . . . . . . .. . . ....nnfian.afl.......s.n..\. . . . 0....l1 . 51.93.“- . .....uuul

“Ed“
{0
9

0
cu,

.n“
3.

.l‘

1... .4:- . .. .. . . ..II
3.. . . rt.-.
...".L l‘

x
.g

.0.“ .. =
; ...... .
“win“
. . . ...». ...
.. . . ...... . «mum....nmv..... ..
. .. . . . . .... u ,e v! . .
. . .. . . .. . .Jflnfimafl luv .
t ... . . ch?‘ 3...! _ .1 .
. . . .a , ...: 5‘. g.‘ .. \u .13. 31'. ..
.. .. ... ..-ufiruwhrnfi £39.... .... . . . . ..
Lam... ......wfie A .5 «a... .. .. . . . . , . .. . . .
...- Ram-Ti.-. 33...}... "33.4%....5. . . . .., . . . . .11....1
1 (Binding. IASVIJ... ...-9...... ...uUaL. ... .. . . .. . , ... . . p ...: .. c0 I‘lsrro.
.-. o 2.212»... . ....lwhrfitu . .. . ... . .. l .. : ... 1.33. 1-..! ..
. ..5nur.VL.v.n.S. . tad, .... . . . x C . . , . .. . gl..u.....l.hL.!.uu..l..
.. 0’ ..-! c......y..~.&.fl.fl4fi....v.uuv . . 11 ... . . . .. .. .. . , . 3.61:“.rfhflusn .
.31... . at- 311.. .v..- u v . .. . . .. . . . . .. . ... .. .nv:..‘..0..v 0.5..-.
a?» ........fiw..a.....fi.mw. . . . .. ._ .. . .. ......I. .. .
. l n I 7'..- ‘ 1 I- v u .
.. .....“mmknfifimfinfiwhkfli , I . .1..- .. : . . . . MW-
.thdwiwn 4,... “Wan o I --.. . . . . .
. ......uufl............}..........é- .
..-...”aé. 3....-. .
mum: la... . . .4
6 xfilHu.
...vhHoN‘uhu £43...-
.. no..- ... .-.! _
.... . r . , ifimmu ...
.11.... .

u-

',

h'
.i!‘

Lu||. I-
wn Minn. . v.
I "Ul .I‘" 0."! f ’ a
wall“)... bk. . ‘5... v: tn 7
. . .....YAfiIMWruy-.-
no- 1.. I. v
.. Liv ‘1 . , ..UI .
001‘. .‘czdr ......
.v . .. cu.

inf-{'2’ ‘

. 1r... . . i
2-..... .lfibzu.
, ...”...R. ...“...s... an ..e. ..
..lvru...n.nvtql v u ..
.9 L135... . ..

‘3. ~' 0. ‘l I?- ‘VC - . .‘
. ....ox . 6... .....f3..u.4...tuufi..urn!tl.rhfl...

. .1--. ......nkwa... 3 .au.n.h....

... 0.5.0..
.....-

. u)l -.VPDDUI
l.-.) ...... . .
394...“? - ....
. .

b. .v..t..dul‘z
....|. . 9.00.1“..va4. ,

Awfiih {uni-...-..» 5.. .. . .lu .....tluni .
|nll llzio . . . (lb. (.vuf l
.i. M. Ld.o...iv.-alnlx.. . . v.ul. .-.s... ‘ Q‘ -.l \«t.. It»...

. . KP $.I‘4Xliitlx . . v rQIIJAINA’IZI‘I.

.K.vl.lu.0|i‘..lk . a . “UNI. 0N5»?
..‘bll..-|Il:n!l‘.v . . .

. -..-31.. .. . . .75 a
tutti ...lt. v}! k..- c- .I.‘
2 v1. .‘Ao....‘.l».u .0. . .
.....l....0.la|..o1 A .
. . I. an...

"71

.ufl»ll .nou “WEN-fivbw
1. .l... I..:.. ....I.-.I.l .----.L ..t . -.. . .
. c. on... ‘w c. . ...> . . trunm- .\ 3.0.“; ..J....tv . . 4.. . . .
I: II.“tE"I.‘\r~u. A u . .v I I .A > ..rl Q V I. Inltl‘ >~)‘-‘
d I‘ I‘UI. hi '0. 1 I'l.ll.l~.'l‘|\‘ n'xlthr l|lt
Lv.nlfir.o¢0 ...- . an: . ...-9......ru0... 1!... 7!
ol-i'IV“.-o.'(u'\‘f~.1.luvllul.o C 'u‘filv‘vt v \‘Ikil q . u
. . .. .||.. A... .1. -..I‘Ot «trolivl .--!{W . .l ..l... n
. . ...|I. . v-1... 3A!“ . ..th v.1.‘5r
. l.

.... .. f o
.‘I.‘.:‘ 1.1L \
. .v-l .vl..t. .|\.

 

Ln...

'7

...: k
1.1..\. . 2 ...

.u .n|,ll0 l-‘n‘.'|.‘.
. lg... v ......urnwdmbft

A II‘XV.I‘III‘COIV.

 

 

 

 

 

 

I»illlllllllll

This is to certify that the

thesis entitled

FACTORS AFFECTING RUTTING OF FLEXIBLE
PAVEMENTS IN PAKISTAN

presented by

Javaid Akbar

has been accepted towards fulfillment
of the requirements for

 

 

 

 

. ivi En ineerin
MS degree 1n C l g g
Major professor
8 21 9
Date / / 7
0-7639 MS U is an Affirmative Action/Equal Opportunity Institution

_¥

l LIBRARY
M'Chlaan State
University

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

FACTORS AFFECTING RUTTING OF FLEXIBLE PAVEMENTS IN
PAKISTAN

By

Javaid Akbar

A THESIS

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

MASTER OF SCIENCE

Department of Civil and Environmental Engineering
1997

ABSTRACT
FACTORS AFFECTING RUTTING OF FLEXIBLE PAVEMENTS IN
PAKISTAN
by
Javaid Akbar

Flexible pavements are designed to withstand structural and functional failures.
Rutting is a structural defect associated with functional implications. Rutting is mainly
caused by wheel loads and accelerated by material and environmental factors. Although
all pavement layers (base, subbase and roadbed soils) contribute to rutting, the
contribution of the Asphalt Concrete (AC) layer alone could be very significant.

The pavement network in Pakistan is experiencing premature rutting problem in
the AC layer. The main objective of this study is to identify the AC mix factors and
physical properties that contribute to pavement rutting.

Fifty pavement sections were selected from a large pool of pavements. For each
selected pavement section rut measurements were made and cores were obtained. The
cores were subjected to various laboratory tests to ascertain the asphalt mix properties.

A multivariate regression analysis was performed to determine the relationship of
rut depths and AC mix properties. The result showed that air voids in the range of 4 to 7
percent, and low penetration of asphalt, can provide significant improvement in rut
resistance of the AC mixes. High loads and the longitudinal grade of pavement contribute

to rutting.

In memory of my parents and to wife, children, sister, brothers and all other loving

humanity.

iii

ACKNOWLEDGMENTS

Gratitude to Almighty Allah who very kindly blessed me with opportunity,
wisdom and strength to undertake and complete this task. The writer wishes to express
his appreciation to his major professor Dr. Gilbert Y. Baladi for his constant guidance and
encouragement during the research. Thanks to the members of the writers master’s thesis
committee.

Dr. W. C. Taylor (Professor)
Dr. Neeraj Buch (Asst. Professor) '

Indebted to Pakistan Army, National University of Science and Technology
Pakistan for sponsoring my study program. Obligations to Dr. Imtiaz Ahmad, Dr.
Shamshad Ahmad Khan, Dr. Tariq Mehmmod, Mr. Noor Muhammad, Bureau of
Statistics, Islamabad (all fiom Pakistan) for their valuable guidance.

Thanks to Military College of Engineering, National Highway Authority, National
Transportation and Research Center, Pakistan, for their cooperation in regard of
laboratory testing and data collection. The help, regarding computer problem shooting,

of Mousa Abbasi (transportation student of MSU) is also acknowledged.

iv

TABLE OF CONTENTS

 

CHAPTER 1
INTRODUCTION ........................................................................................................ . 1
1.1 GENERAL ............................................................................................................ 1
1.2 PROBLEM STATEMENT ................................................................................... 2
1.3 RESEARCH OBJECTIVE ................................................................................... 3
1.4 SCOPE ................................................................................................................ .4
CHAPTER 2
LITERATURE REVIEW ............................................................................................................... .9
2.1 GENERAL .......................................................................................................... .9
2.2 RUTTING ......................................................................................................... .11
2.2.1 Mechanism .............................................................................................. 14
2.2.2 Factors Affecting Pavement Rutting ....................................................... 19

2.2.2.1 Tire Inflation and Tire-Pavement Contact Pressures........... . . ...19

2.2.2.2 Consolidation and Field Compaction ...................................... .23
2.2.2.3 Aggregates ................................................................................ 27
2.2.2.4 Sand and Mineral Filler .......................................................... .34
2.2.2.5 Asphalt Type and Content ...................................................... 35
2.2.2.6 Environmental Factors .............................................................. 41
2.2.2.7 Ranking of Factors .................................................................... 42
2.2.2.8 Rut Prediction Models ...................................................................... .43

CHAPTER 3

 

RESEARCH PLAN ........................................................................................................................ 55

3.1 GENERAL ....................................................................................................... .55

3.2 RESEARCH PLAN ......................................................................................... .56

3.2.1 Pavement Selection and Investigation ................................................... 58
CHAPTER 4

FIELD AND LABORATORY INVESTIGATION ............................................................ 60

4.1 FIELD INVESTIGATIONS ............................................................................ 60

4.1.1 Pavement Selection............... ............................................................... 60

4.1.2 Measurements of Rut Depths ................................................................ 65

4.1.3 Core Locations ..................................................................................... .65

4.1.4 Marking, Coding and Coring ................................................................. 66

4.1.5 Traffic Volume and Loads ..................................................................... 67

4.1.5.1 Definition - Truck Factor .......................................................... .68

4.1.5.2 Service Life ............................................................................... .68

4.2 LABORATORY INVESTIGATIONS ............................................................. .68

9 4.2.1 Bulk Specific Gravity Tests (ASTM-D2726-90) .................................. .72

4.2.1.1 Definition ................................................................................... 72

4.2.1.2 Summary of Test Method .......................................................... 73

4.2.1.3 Significance and Use ................................................................ .73

vi

4.2.1.5 Sampling .................................................................................... 73

 

4.2.1.6 Procedure ................................................................................... 74
4.2.1.7 Calculations ............................................................................... 75
4.3 EXTRACTION TESTS ..................................................................................... 76
4.4 RECOVERED ASPHALT PENETRATION ................................................... .77
4.5 SIEVE ANALYSIS .......................................................................................... .77
4.6 AGGREGATE ANGULARITY ...................................................................... .78
CHAPTER 5
ANALYSIS AND DISCUSSION .............................................................................................. .87
5.1 GENERAL ........................................ .............................................................. .87
5.2 STATISTICAL ANALYSIS TECHNIQUES AND ISSUES ......................... .91
5.3 ANALYSIS AND DISCUSSION OF RUT DATA ........................................ .97
5.3.1 Test Results .......................................................................................... .99
5.3.2 Determination of Material Properties for a Core Location ................... 99

5.3.3 Statistical Analysrs 100

 

5.3.4 Regression Equation ........................................................................... .111
5.3.5 Sensitivity Analysis and Engineering Interpretation ........................... 119
CHAPTER 6
CONCLUSIONS AND RECOMMENDATIONS ................................................ 129
6.1 Summary ...................................................................... 129
6.2 Conclusions .................................................................. 129

vii

6.3 Recommendations .......................................................... 1 30

REFERENCES .................................................................................................... 1 32

viii

Table 1.1 :

Table 2.1 :

Table 2.2 :

Table 2.3 :

Table 2.4 :

Table 2.5 :

Table 2.6 :

Table 4.1 :

Table 4.2 :

Table 4.3 :

Table 4.5 :

Table 4.4 :

Table 5.1 :

Table 5.2 :

Table 5.3 :

Table 5.4 :

Table 5.5 :

Table 5.6 :

LIST OF TABLES

Experimental Design Matrix ........................................... 7

Average rut (percent total rut) in each layer of section 51

of the AASHO road test (7) ........................................... 18
Percent rutting in various layers due to distortion section

51 AASHO road test, 1960, (7) ........................................ 18
Effects of tire type and tire inflation pressure on the

maximum tire pavement contact pressure (16) ...................... 22
Summary of mix densities (lbs/c ft) (31) ............................. 33
Effects of asphalt cement factors on pavement distress(40) ....... 41

Summary of ranking of mix properties related to

pavement performance (11) ............................................ 45
Selected flexible sections ............................................... 65
Truck Factors at Taxila (63, 65) ....................................... 72
Field data all test sites .................................................. 73

Results of extraction, penetration and bulk specific
gravity test for AC course of pavement sections .................... 82

Results of sieve analysis

for AC course ............................................................ 87
Field and laboratory data for test samples .......................... 105
Descriptive statistics of table 5.1 ..................................... 109
Correlation matrix for extracted core locations .................. 11 1
Correlation matrix for final variables ............................... 117
Model summary ........................................................ 117
Collinearity diagnostics ............................................... 118

ix

Table 5.7 : Results of regression analysis and collinearity

diagnostics TOL and VIF ............................................. 120
Table 5.8 : Regression matrix for depths of rutting ............................. 122
Table 5.9 : Residual statistics of equation 5.4 .................................... 122

Figure 1.1 :
Figure 1.2 :

Figure 1.3 :

Figure 1.4 :

Figure 2.1 :

Figure 2.2 :

Figure 2.3 :

Figure 2.4 :

Figure 2.5 :

Figure 2.6 :

Figure 2.7 :

Figure 2.8 :

Figure 2.9 :
Figure 2.10 :

Figure 2.11:

LIST OF FIGURES

Map of Islamic Republic of Pakistan ................................. 5
Detailed location map of areas included in the study ................ 6

A schematic representation of a pavement test section
with three test sites ........................................................ 8

A schematic representation of a pavement test locations
along 100 ft long test sites ............................................... 8

Illustration of critical stress / strain in typical pavement
structure (4) .............................................................. 13

Study of transverse profile of loop 4 and 6 of the
AASHO road test (7) ................................................... 17

Variation of the compressive strain at the top of the
subgrade for different thickness’ and tire pressure (14) ............ 25

Variation of the compressive strain at the top of the
subgrade for different thickness’ and two levels of load (14) ..... 26

Pavement subgrade rutting damage life for thickness’
and loads ................................................................. 26

Rutting as a function of voids in the total mix (V TM) (9) ........ 27

Model for rut development in AC. Pavement performing

as designed (10) ......................................................... 29
Model for rut development in AC. Pavement performing
excessively (10) ......................................................... 29
Relationship between rut depth and stability (20) .................. 31
Effects of aggregate top size upon percent air voids

in mineral aggregate (31) .............................................. 34
Effects of aggregate top size upon percent air voids

in the mixes (31) ........................................................ 34

xi

Figure 2.12 :

Figure 2.13 :

Figure 3.1 :
Figure 5.1 :

Figure 5.2 :

Figure 5.3 :
Figure 5.4 :
Figure 5.5 :

Figure 5.6 :

Binder contribution to different distresses (4) ...................... 38

Permanent deformation as a function of fines (Brampton

road test section 3 ) (9) ................................................. 49
Overall research plan ................................................. 59
Observed verses the predicted rut depths ........................... 124

Percent difference between the observed and
the predicted rut depths ............................................... 125

Rut depths as a function of air voids ................................ 129
Rut depths as a function of recovered asphalt penetration ....... 130
Rut depths as a function of 18 kip cumulative ESALs ........... 131

Rut depths as a function of gradient of road ....................... 132

xii

CHAPTER 1

INTRODUCTION

1.1 GENERAL

The structural design of asphalt concrete (AC) pavements and AC overlays has been
an evolutionary process based primarily on experience and judgment of the highway
engineers, augmented by empirical relationships developed through research and field
observations. A proper design of AC pavements requires the consideration of several
complex and interrelated factors such as, the AC mix design, environmental factors and
engineering properties of various pavement layers. Recently efforts have been made to
analyze these factors to enhance existing empirical design procedures and to develop
rational design methods/models based on theoretical background such as the theories of
elasticity, visco-elasticity and plasticity. Today, design methods for flexible pavements and
overlays could be divided into two groups, empirical and mechanistic—empirical. The main
design consideration in both groups is to limit the vertical compressive strain induced at the
top of the AC layer, base and roadbed soil to control permanent deformation and to
minimize the tensile strain induced at the bottom of the AC or the stabilized layers to

control fatigue cracking.

2

Both rutting and fatigue cracking are load associated distresses and are affected by
several other factors such as traffic volume, load, gradient of road section(which reduces
speed of trucks), material properties, construction quality, layer thicknesses and
environments. Therefore, any methodology based on empirical or mechanistic approaches,
must model these factors in order to predict the pavement behavior efficiently.

Rutting develops through different mechanisms namely repetitive plastic
deformation (no volume change) and densification (volume change). The rate and amount
of rutting resulting from repetitive plastic deformation, is a function of the load magnitude,
load repetitions, material properties, construction quality, and temperature. Prevention of
this type of rutting is handled through pavement structural/material design. Rutting caused
by densification can be controlled by better quality control, compaction specifications, and
enhanced design.

The principal objective in design of AC pavement structures is to reduce both
rutting and fatigue cracking potentials. This objective can be achieved by using a balanced
design procedure, a good quality asphalt mix, better specifications and quality control

practices.

1.2 PROBLEM STATEMENT

In Pakistan, permanent deformation is the principal distress mode in flexible
pavements. Several factors affecting this type of distress have been identified including:
1. High truck loads/volumes and high tire pressure.

2. The asphalt mix properties.

3
3. Inadequate pavement structural design.

4. Inadequate construction practices (e.g. compaction, segregation).

5. Softening of the asphalt binder due to high temperatures.

6. Strength and type of the materials used in the various pavement layers.
7. Softening of the base, subbase, and roadbed materials due to moisture infiltration in
some regions of Pakistan.

Existing rehabilitation methods to correct rutting of asphalt pavements in Pakistan
generally involve cold milling followed by AC overlays. The rehabilitation process is found
costly and provides no assurance that the overlaid pavement will not rut again. In many
instances, newly constructed and/or rehabilitated asphalt pavements have experienced
rutting in a short time period after construction. Hence the need to identify the fundamental
causes of rutting and to determine the remedial measures have been recognized.

1.3 RESEARCH OBJECTIVES

The objectives of this research study include:

1. Analysis of the asphalt mix properties (i.e. percent aggregate, sand and fine

contents, aggregate angularity, binder type and content, and the percent air voids)

that affect pavement rutting.

2. Carryout field investigations comprising of measurements of rut depth and
road gradients and computation of traffic loads(ESALs).

3. Develop a statistical rut model and identify the variables that influence
pavement rutting in Pakistan (i.e. asphalt content, pavement thickness,
recovered penetration of asphalt, fines and sand content, aggregate type

and content, air voids, AC layer thickness, gradient of road and ESALs etc.

4

4. Prioritizing the factors in order of their significance on rutting, make

conclusions and recommendations to improve the existing asphalt mix

design and pavement construction practices.
1.4 SCOPE

To accomplish the stated objectives, 54 flexible pavement sections, were selected
for this study from SAHIWAL (km-1102 distance measured from Karachi) to PESHAWER
(km-1700) areas by physically surveying the routes. The criteria used in selection was that
the selected sections should represent the spectrum of pavement cross sections, paving
material, traffic volume and loads, found in most of the areas of Pakistan and exposed to
similar weather conditions(hot weather). Detailed locations can be viewed fiom figures 1.1
and 1.2. Full factorial experiment matrix (Table 1.1) had been formed. Volume of
traffic(yearly ESALs) and ranges of AC thickness have been considered as variables for this
task. All cells of the matrix were studied depending upon availability of pavement sections
within desired parameters of dependent variable (Rut). Each test section of varying length(1
to 5 km), has test sites of 30 meter long with varying levels of distress severity. Distress
levels were measured at five test locations(6 meter apart) within one test site, as per
AASHTO instructions(2) and then the average was taken (Figures 1.3 and 1.4). The study
encompasses a synthesis of available information on rutting, measurements of rut depths
and road gradients, extraction of representative core samples and laboratory testing of these
samples to include, measurement of AC thickness, calculation of bulk specific gravity/air
voids, recovered asphalt penetration, percent asphalt content, aggregate, sand and fines.
The test data has been analyzed, statistical rut model developed and sensitivity analysis has

been under taken to offer conclusions and recommendations

 

 

r. /-. c mi. /

)
"run—-..“...r -... _ .... “...—-...

I

 

 

Figure 1.1. Map of the Islamic Republic of Pakistan

 

 

    
    
     

l

l

PESHAWAR l
l

l

l

ISLAMABAD l
l

l

l

i

‘ l
l

l

l

l

GUJRAT l

l
GUJRANWALAi

l

l

l

i

i

l

l

l

l

LAHORE l

l

l

l

i

l

 

Figure 1.2: Detailed location map of areas included in the study.

Table 1.1: Experimental Design Matrix.

 

YEARLY ESALS (millions)

 

0-2 2-4 >4

 

 

 

ASPHALT 10 - 18 2 5 8
CONCRETE
THICKNESS
(cm)
> 18 3 6 9

 

 

 

 

 

 

* Including asphalt wearing course and asphalt base course

 

Test Site (30 Variable Distance Test Site Variable Distance Test Site
(30 M) (30 M) (30 M)

 

 

 

 

 

 

 

 

Figure 1.3: A schematic representation of a pavement test section with
three test sites.

 

6M 6M 6M 6M 6M

 

 

 

 

 

 

 

 

X = Test location

Figure 1.4 : A schematic representation of the five test locations along
30 meter long test site.

CHAPTER 2

LITERATURE REVIEW

2.1 GENERAL

The load carrying capacity of flexible pavements is brought about by the load-
distributing characteristics of their layered systems (1). In general, flexible pavements
consist of a series of layers with the highest quality material placed at or near the surface.
Hence, their strengths are the result of building up thick layers and thereby distributing the
load over the roadbed soil (1 ).

Two types of load related distress (rut and fatigue) are most commonly found in
flexible pavements. Rut is a surface depression in the wheel paths. Pavements uplift may
occur along the sides of the rut; however, in many instances, ruts are noticeable only after
rainfall, when the wheel paths are filled with water. Rutting stems from a permanent
deformation in any of the pavement layers or subgrade, usually caused by consolidation or
lateral movement of materials due to traffic loads. Rutting may be Caused by plastic
movement in the AC mix due to high atmospheric temperatures or inadequate compaction
during construction. Significant rutting can lead to major structural failure of the pavement
and hydroplaning potential. Wear of the surface in the wheel paths fiom studded tires can
also cause a type of "rutting" (2).

It should be noted that the contribution of the AC layer to the total pavement rutting
due to densification is negligible since this layer is typically compacted to near its
theoretical maximum density during construction. Permanent deformation in the AC layer
is mainly the result of lateral distortion due to repeated shear deformation (3).

Existing flexible pavement design methods can be divided into two categories;
empirical and mechanistic-empirical. Most of the empirical design methods are based on

statistical equations derived from field observations of pavement rutting and surface

9

10

roughness. Mechanistic-empirical design methods, on the other hand, are mainly based on

two criteria:

1.

Minimizing the rut potential of each pavement layer by limiting the magnitude of
the compressive stress induced at the top of that layer by a moving wheel load.
Maximizing the fatigue life of the AC layer by minimizing the induced tensile
stress at the bottom of the layer due to a moving wheel load.

Regardless of the pavement design method (empirical or mechanistic—empirical)

employed in the design of flexible pavements, the design process involves two major steps

as follows:

1.

The design of the asphalt mix which involves the proportioning of the different
ingredients (coarse and fine aggregates, mineral fillers, and asphalt cement) in the
mix and the compaction effort.

The thickness design of the AC course and the other pavement layers (base and
subbase), which involves the evaluation of the behavior of these layers under the
anticipated traffic load and environmental conditions. Since the outcomes of the
asphalt mix design process (step 1) affects the engineering properties of the mix,
the thickness design of the pavement, and the pavement performance, the two steps
(mix design and thickness design) and construction practices must be considered in
a comprehensive way so that the desired pavement performance is assured.

Rut potentials of pavement can be minimized by taking balanced engineering steps

during the material design (asphalt mix design), pavement design process and construction.

These include (4):

l.

Engineered asphalt mix design that can withstand the expected traffic loading
without plastic yielding, resists the induced tensile stress without cracking, and

resists low temperature cracking.

ll
2. Balanced pavement design process that provides balanced and adequate layer

stiffness and thickness to reduce the induced compressive stresses at top of the
base, subbase layers, and roadbed soil, which cause plastic deformation (rutting) of
these layers and to minimizes the tensile stress and strain induced at the bottom of
the AC thereby increasing its fatigue life (see figure 2.1) (5).

3. Good construction practices that deliver adequate and uniform compaction of the
various layers.

Existing information related to rutting is reviewed in succeeding paragraphs.

2.2 RUTTING

Rutting (permanent deformation) is a load related distress ( caused by high axle
loads and volumes ) and is accelerated by material, environmental factors and by inadequate
construction practices. Rut is the accumulation of plastic (permanent) deformation in the
roadbed soil, in base and subbase layers due to compaction and/or consolidation under
traffic loading. Rutting of asphalt layer may be caused by combination of volume change
and shear deformation resulting from the application of traffic loads. In general, the rut
potential can be minimized by using good quality materials, coupled with adequate
pavement design and construction practices.

Rutting occurs as a combination of densification (volume change) and shear
deformation. It has been observed by a number of investigators that the shear portion is
more significant than the volumetric one. The major portion of permanent deformation

occurs during initial loading (6).

l2

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

C9 G)

_-—__-----—

. ..--...____-._- __ _.
r
.L,,I a

 

 

 

 

 

 

Figure 2.]: Illustration of critical stress/strain locations in a typical

flexible pavement structure (4).

13

The initially soft material is being load-compacted and stiffened until the yield
point is reached where additional plastic shearing accompanied by dilation (increase in
volume due to the rearrangement of particles) occurs. The irreversible deformation
during the entire process is strongly time dependent, because of the viscous nature of the
bituminous binder. The factors which influence the rutting behavior

include (6) :

aggregate gradation, size, angularity and surface roughness;

mixture composition;

compaction procedure;

temperature;

load magnitude, distribution and frequency;

thickness of the asphalt concrete layer;

Air voids in the AC mix; and

longitudinal grade of the pavement.

For all granular materials (and for that matter, for most other materials), the distinction
between volumetric and deviatoric deformation is essential, once the linear threshold has
been exceeded. For some stress levels, both volumetric and deviatoric deformation have
their reversible and irreversible(permanent) portions. Each of these may or may not
depend on the duration of the load and is modeled accordingly as either time-dependent
or instantaneous . In addition, volumetric deformation Can be such that volumetric stress
performs positive or negative work. The latter is termed as dilation. The volumetric

irreversible deformation due to volumetric stress occurs not only for high stress but also

l4

initially which is referred to as densification - an irreversible initial volume change.

The major portion of the irreversible deformation is due to shear stresses in the
uncompacted material, which suggests that better compaction of the asphalt layer should
significantly improve rutting behavior. Also low asphalt content is shown to be
beneficial (6).

For any flexible pavement section , its rut is typically determined by computing the
mean rut depth (mean difference in elevation between the rutted wheel paths and the
surrounding) of the pavement surface along the length of the project. Based on the mean
rut depth, three severity levels can be defined as follows (2):

1. Low -Mean rut depth of .25 to .5 inch ( 6 to 12 mm ).
2. Medium -Mean rut depth of .5 to 1.0 inch ( 12 to 25 mm ).
3. High- Mean rut depth of more than 1.0 inch ( 25 mm ).

To determine the mean rut depth, a straight edge of 4 feet long, should be laid across

the rut and the maximum depth is measured. The mean depth should be computed from

measurements taken at every 20 feet interval along the length of the rut (2).

2.2.1 Mechanism

As mentioned in foregoing paragraphs, pavement rutting is the sum of the rut in the
AC, base and subbase layers and in the roadbed soil. Figure 2.2 shows the results of a study
of the transverse profile of loops 4 and 6 of the AASHO Road Test (7). It can be seen that
rutting had taken place in all pavement layers and in the roadbed soil. The contribution of
each layer to the total pavement rut varies from one pavement to another. For example, the

average rut in each layer as a percent of the total pavement rut of section 51 of the AASHO

15

Road Test, is shown in Table 2.1. In general, a narrow rut channel indicates that the
primary rut is in the AC layer. The wider the rut channel, the deeper is the seat of rutting.

Table 2.2 provides a list of the relative percentages of the permanent deformation of
section 51 of the AASHO Road Test that are attributable to distortion alone. It can be seen
that, in general, permanent deformation is largely due to lateral distortion.

Based on laboratory data, Morris (3) concluded that densification and lateral
distortion in compacted asphalt concrete specimens are largely a function of the test
conditions. He added that, in the field, asphalt pavements are subjected to densification in
the compression zone and to lateral distortion in the tension zone. Tests on asphalt concrete
used at the Brampton Test Road (9), confirmed the above observations. It was noted that
the rut potential in the tension zone in asphalt concrete pavements is higher than that in the
compression zone because of distortion.

Rutting occurs by virtue of consolidation, compaction, or lateral distortion of the
pavement due to repeated traffic loads. Several factors affect the magnitude of rut and its
time rate of accumulation. These include (5):

1. Construction factors including inadequate compaction (either low compaction efl°ort
or compaction at lower temperatures than those specified).
2. Asphalt mix factors that include soft (low viscosity or high penetration) asphalt

cement, high air voids, rounded aggregate, and excess sand in the mix.

3. Environmental factors that include high temperatures which soften the AC layers,
and high moisture content or saturation of the lower layers (base and subbase) due

to inadequate drainage.

 

 

 

F . ...... . .

    

    

 

 

  
    

 

 
 
  

       

1 i
o 2

 

Loop 6 6-3-l2 Design

e Axle Load 1
|.97'__—-—---—-—-_-———--__—-_ T _________

7 “"

l
a

 

 

 

 

  
      
   
 

% 7

 

 

 

22,.- m\
/

° N

@U

l
4
L

 

cmc m 1
/ ASE ign
PATCH Axle Loo

......

N

o 4es

fl

nde

 

~ -— . —“ —————
-- M: T_ _ ‘ .... —.— _E ——————— i
m «. . 1:233:14, 11.». $._ act-v. . ----------------
5 l— ' [2.15% s»- -.—— - — . j
. 5_ ' ~

1: W 414' V 1:"; 3:?Wfi H —i—
flay/>2» Wj

1 1 I 4 1

4 s B
rrrrrrrrrrrrrrrrrrrrrr

Figure 2.2 : Study of the transverse profile of loops 4 and 6 of the AASHO ROAD

TEST (7).

l7

 

 

 

 

 

 

 

 

 

 

 

 

 

Table 2.1 : Average rut (percent of total rut) in each layer of section 51 of the AASHO Road Test (7).
Percent of the total rut.
AC Base Subbase Roadbed soil
32 14 45 9
Table 2.2 : Percent rutting in various layers due to distortion, section 51 of AASHO Road Test,
1960, (7).
MATERIAL % Rutting due to distortion during a AVERAGE
given season
SPRING SUMMER FALL
A/C SURFACE 82 82 76 80
BASE O 70 NO Figs.inconsistent
MEASURE.
SUBBASE 97 96 NO 95.5
MEASURE.

 

 

 

 

 

 

 

 

l8
4. Tire factors such as studded tires and high tire pressures.

Available information related to these and other factors affecting pavement rutting is
presented in the next subsections.

Analysis of data from field test sites in Alabama indicate that permanent
deformation causing rutting is generally confined to the top 3 to 4 inches (surface and
binder courses). There was little evidence that lower base/subbase courses or subgrade
were significant contributors to rutting (10).

Permanent deformation is affected by the degree of compaction of the asphalt
concrete mix. Two types of permanent deformation occur - consolidation and lateral
distortion. Consolidation deformation, which is a relatively minor problem, comes about
when traffic reduces the air void content of the mix from the value at the time of
construction to a minimum air void content which generally occurs after several years of
traffic. Permanent deformation which is due to lateral distortion is primarily a function of
the properties of the mix. A tender mix, compacted to a low air void content will
undergoes less permanent deformation than if it is compacted to higher air void contents.
Lateral distortion can also follow traffic consolidation if the mix densifies down to a
critical air void content (normally less than 3 percent) and in fact becomes over lubricated
with asphalt cement. Environmental factors that allow the compaction equipment to
obtain density with less compactive effort thus affect both types of permanent
deformation (1 1).

Kang W. Lee and Mohammad A. Al-Dhalaan (12), concluded from road test
carried out in Saudi Arabia that the rut damage was limited to the wearing course; that is,

the subgrade and the base course are intact and the major cause of failure was the

l9

inadequacy of the specified mix for the loading and environmental conditions, for
example, high temperature.

2.2.2 Factors Affecting Pavement Rutting

2.2.2.1 Tire Inflation and Tire Pavement Contact Pressures

In U S A as well as in Pakistan, asphalt surfaced pavements are experiencing premature
rutting due to increased traffic volume, loads and/or increased truck tire pressure. Surveys
in the States of Illinois and Texas indicate that the tire pressures have increased
substantially over the last few decades. An average tire pressure of 96 psi with a maximum
of 130 psi were recorded in the Illinois survey. The Texas survey showed an average tire
pressure of 110 psi with a maximum pressure of 155 psi (13). Similar study revealed that
the tire pressure in PAKISTAN ranges from 95 to 145 psi (15).

Typically, the rut potential of asphalt pavements has been evaluated on the basis of
the magnitude of the compressive stresses induced at the top of the base layer and roadbed
soil due to an 18-kip (80 KN ) single axle load and a constant tire pressure (typically 85
psi). Experimental studies conducted by GoodYear tires and Rubber Company indicate
that (l 6):

1. For a constant tire load, increasing the tire inflation pressure causes a shift in the
point of maximum contact pressure to the center region of the contact area between
the tire and the pavement surface.

2. Regardless of the tire load and tire pressure, the tire-pavement contact pressure is
not uniform within the tire-pavement contact area. The distribution of the contact
pressure is a function of the tire type and design. For example, contact pressures as
high as twice the tire inflation pressure were measured for three tire types as shown
in Table 2.3.

Smith and Bonquist (17) studied the influence of tire type and tire inflation pressure

on pavement performance. They conducted full-scale pavement tests using the Federal

20

Highway Administration (F HWA) Accelerated Loading Facility (ALF) located at Turner

Fairbank Highway Research station in Mclean, Virginia. The study examined the effects of

tire pressure, tire load, pavement cross-section, and environmental conditions on pavement

response (stresses and strains) and on pavement performance (rut potential and fatigue life).

They made the following conclusions:

1.

The effect of wheel load on pavement responses is greater than the effect of tire
pressure. The measured pavement responses (stresses and strains) doubled for an
increase in load from 9,400 pounds to 19,000 pounds, while increasing the tire
inflation pressure from 76 psi to 140 psi resulted in a less than 10 percent increase
in the measured response. This conclusion supports the results of mechanistic
analysis of flexible pavement structures. For example, Baladi used MICHPAVE
(linear/nonlinear finite element computer program) to analyze the stresses and
strains induced in the pavement layers due to various wheel loads and tire inflation
pressures. He reported that the effects of increasing tire pressure on the induced
stresses and strains in the pavement are much smaller than those due to increasing
wheel load.

The effects of tire pressure and/or wheel load on pavement rutting are much higher
for thin pavement sections (less than 2-inch AC surface) than for thick sections
(more than 4-inch thick AC surface).

Further, higher temperatures cause higher rut potential. Hence, the combinations of

high tire pressure, high wheel load, high temperatures, and thin pavement sections are

extremely damaging to flexible pavements. Although, the average truck tire inflation

pressure has increased by about 20 psi over the last thirty-year period, the above findings

suggest that this increase has an insignificant effect on pavement rutting (4).

21

Table 2.3 : Effect of tire type and tire inflation pressure on the
maximum tire - pavement contact pressure, (16).

 

 

 

 

 

 

 

 

 

 

Tire type Tire inflation Max. contact Ratio (MCP/TIP)
and load pressure (TIP) pressure (MCP)
(Psi) (Psi)
11-24.5 65 117 1.8
4250 lbs 75 114 1.5
95 118 1.2
llr24.5 75 122 1.6
4250 lbs 95 152 1.6
115 182 1.6
385/65R22.5 100 200 2
120 226 1.9
140 252 1.8

 

 

 

 

 

 

22

On the other hand, the reported increase in wheel load(e.g., the average wheel-load in
Michigan and in Pakistan, has increased by 4000 and 8000 pounds respectively over the last
twenty-year period) has significant effect on the rut potential (4 , 8 ).

Chen (14) employed the three dimensional Texas Grain Analysis (TEXGAP-3D)
program (a 3-D linear finite element computer program) to study the effects of various
wheel loads, nonuniform tire pressures, and nonuniform tire-pavement contact pressures on
the induced stresses and strains and the resulting damage in asphalt concrete pavements.
Results of the analysis were also compared with those of the ELSYMS computer program
which uses a circular contact area with uniform contact pressure distribution. Chen utilized
a rutting model developed from the AASHO Road Test data to analyze pavement rutting.
He made the following observations:

For a constant axle load of 4500 pounds, a 47 percent increase in the tire inflation
pressure produces less than 2 percent increase in the compressive strain developed
at the top of the subgrade as shown in Figure 2.3.

2. For a constant inflation pressure, a 20 percent increase in the axle load produces
approximately a 20 percent increase in the subgrade compressive strain at the top of
the subgrade as shown in Figure 2.4.

3. For the same tire inflation pressure, increase in axle load from 4500 pounds to 5400
pounds results in a 50 percent reduction in pavement life based on subgrade
compressive strain as shown in Figure 2.5.

Based on these observations Chen et a1. concluded that the axle load, not the

inflation pressure has a major influence on subgrade rutting.

23
2.2.2.2 Consolidation and Field Compaction

Proper specifications and quality assurance regarding asphalt mixture composition

and compaction decrease the rutting potential due to densification.

Barksda1e(1 8) cited the work of Raithby, Epps and Acott, and reported that asphalt
mixes placed in the field at a compaction level equal to the 50-blows Marshall compaction
experienced gradual compaction and densification due to traffic loading (18).

Huges and Maupin(l 3) monitored the rut depth of asphalt pavements made by using
different asphalt mixes (see Figure 2.6). They measured the rut depth right after
construction, six months, and one year after construction. They reported that mix 1 with the
highest initial voids in the total mix (V TM) has rutted more quickly than Mix 4 which had
the lowest initial VTM. This supports the widely held belief that improving density during
construction reduces rut depth.

Miller (20), based on results obtained from 24 test sites in Arkansas, stated that
asphalt mixtures placed and compacted at air voids between 2.5 to 5 percent provide an
acceptable level of ruts. He added that pavements with deep rutting have air void contents
of less than 1.0 percent (high asphalt content). It should be noted that to prevent bleeding,
the air voids in the asphalt mix should be higher than 2 percent (18).

Brown (22) stated that air void contents of more than 3 percent are needed to
decrease the probability of premature rutting throughout the life of the pavement, whereas,
air voids of less than 3 percent greatly increase the probability of premature rutting.

Huber and Heiman (8) reached a similar conclusion and suggested that asphalt

mixes should be compacted above a threshold value of 4 percent air voids.

24

 

9’
tn

0

 

AC thlckness (Inches)
'0
UI

 

 

 

 

1 .5 - ‘
300 350 400 450 500 550 600 650

Compressive strain top of subgrade (microstrsin)

Figure 2.3 : Variation of compressive strain at the tap of the subgrade for difi'erent

thickness and tire pressure, (14).

25

 

4 \ 2" :
\ \ \
\ \ \\—n—— Nonuesoed hrs
3 5
’u?
g .———— Axie load a 5400 pounds
3 3
I
g 4500 pomds
C
s
'E. 2.5 Treaded me
O
<
4500 pounds
2
5400 pounds

 

 

1.5 . -
300 350 400 450 500 550 600 650 700 750 800
Compressive strain top of subgrade (microstrsin)

 

Figure 2.4 : Variation of compressive strain at the top of the subgrade for different
thickness and two level of loads (14).

 

9’
on

 

Q

P
on

AC thickness (Inches)

 

 

 

 

o 5 10 15 20 25 30 35 40 45 50
Subgtsde mtting (millions oi cycies)

Figure 2.5 : Pavement subgrade rutting damage life for difl'erent thickness and
loads (14).

26

0.15

 

MIX 3

0.1 '

MIX 2

 

 

 

 

5.. 0.05 r l l
V
‘3 \
E
'3 0
I
-0.05 *-
-0.l ' ' '
5 6 7 8

WM i
Figure 2.6 : Rutting as a function of the voids in the total mix (VTM) (9).

27

Lateral distortion can also follow traffic consolidation if the mix densifies down
to a critical air void content (normally less than 3 percent) and in effect becomes
overlublicated with asphalt cement (11).

A study carried out in Albama revealed that in properly designed mixes
densification stabilizes at about 4% air voids and rut depth development ceases or
decreases to very low rates as illustrated in figure 2.7. At about 4% air voids, the rut
resistance in properly designed mixes is optimum. At this stage it is critical that the
aggregate skeletal structure have the ability to resist further densification, and this is best
accomplished with well graded aggregate with angular rough textured particles. Asphalt
content is also critical as the mix reaches about 4% air voids. For pavements that
experience severe rutting, densification continues and second phase conditions develop.
When the air voids reach about 2%, the mix becomes very unstable and plastic flow
deve10ps, as illustrated in figure 2.8 (10).
2.2.2.3 Aggregate

The performance of asphalt mixes depends on providing adequate aggregate
interlock for resisting and distributing the wheel load rather than on the shear strength of the
asphalt cement. Therefore, the size, shape and angularity, and quality of the aggregates play
an important role in the performance of the AC mix. Hence, every precaution should be
taken to insure that a face to face aggregate contact and aggregate interlocking are provided
for the high quality AC mixes. The maximum aggregate size, aggregate shape, and
angularity, and the percent aggregate content in the mix influence the rut potential of

asphalt pavements and Marshall stability and flow of the asphalt mix. Higher aggregate

28

 

 

 

 

 
    

 

I
In: DENSIFICATION
LIJ
Q
'—
D
01 voms
8-4%
#:—
TIME
Figure 2.7 : Model for rut development in asphalt concrete. Pavement performing
as designed (10).
A
, DENSIFICATION PLASTIC
FLOW

I
[—
0.
1.1.1
0 4’
I...
D
0: ' _

vonos { , vouos vonos

8-—4°/.i 4--2°/. 2%-+

. >-

 

 

TIME

Figure 2.8 : Model for rut development in asphalt concrete. Pavement rutting
excessively (10).

29
angularity produces higher Marshall stability, lower flow, and lower pavement rutting (20,

23, 25, 26). Figure 2.9 depicts the rut depth as a function of Marshall stability. Larger top
size aggregate in the asphalt mix, on the other hand, produces higher stability, higher
resilient modulus, higher compressive strength and lower rut potential (28, 29).

Herrin and Goetz (30) studied the effects of the percent aggregate in the mix and
aggregate shape on the stability of the mix. They observed that as the percent aggregate in
the mix decreases, its influence on the stability of the mix decreases. This may be explained
by the fact that as the percent aggregate decreases and the percent sand increases, face to
face aggregate contact is lost producing a ball bearing effect. That is higher percent sand
contents cause the aggregate particles to float in the sand matrix thereby, reducing friction
and stability.

Indeed the European Stone Mastic Asphalt (SMA) mix consists of 70 to 80 percent
crushed aggregate. The SMA is well known for its low rutting potential. In addition, for a
constant asphalt content, large size aggregates cause an increase in the density of the
compacted asphalt mixtures (31). The increase in density is more pronounced when the
aggregate top size is increased from 3/4 to 2-inch as it can be seen from Table 2.4. Further,
increasing the aggregate top size in a mix causes an increase in the percent air void and in
the percent void in the mineral aggregates (V MA) as shown in Figure 2.10 and 2.11 (31).
Similarly, based on literature review conducted by Button at el. (32), he concluded that the
rutting problem can be addressed by using large top size aggregates (1 to 1‘/2-inch),
increasing the specified percent void in mineral aggregate (14 to 15 percent minimum),
replacing most or all the natural sand in the mix with manufactured angular particles,
increasing the minimum allowable air void in the laboratory compacted mix to 4 percent

and limiting the ratio by weight of mineral filler to bitumen to about 1.2.

RUT DEPTH - 1/32 inch

30

40--

351-

Best Fitted Equation
Log Rut = 1598000496 STAB
R2 = 0.49, RMSE = 0.180

30"

25 ~-

20-»

15-1-

IO .-

 

l

 

 

O
lb
m-
I.
d

80 100 120 140 160 180 200
JMD MARSHALL STABILITY - 100 lbs

Figure 2.9: Relationship between rut depth and stability (20).

31
Epps and Monismith (21) reported that crushed (angular) aggregates produce AC

mixtures with low rut potential. Typical requirement for crushed coarse aggregate are:
1. A minimum of 75 percent of particles with two crushed faces.
2. A minimum of 90 percent of particles with one crushed face.

The analysis also indicates that rutting varies geographically and that this
variation can be explained by quality of locally available aggregate. Those areas with
crushed stone and angular natural sands are less susceptible to rutting (10). Changes in
aggregate properties will likely be most fruitful in improving rut resistance.

Button and Perdomo concluded that the round shape and smooth texture of natural
(uncrushed) aggregate particles was one of the chief mixture deficiencies that contribute
to rutting. Aggregate comprises over 90% of asphalt concrete mixtures and provides the
basic load carrying skeletal structure. Well graded angular aggregate, proper selection of
asphalt properties and asphalt contents can produce mixes that are not only rut resistant,
but also durable and resistant to fatigue and thermal cracking (10). Coarse aggregate
retained on the 4.75-mm (No.4) sieve should continue to be at least 85 percent of
particles with two or more fractured faces for wearing and binder courses (32).

The Michigan Department of Transportation (MDOT) has developed a special
provision for designing the asphalt paving mixtures for heavy duty pavements to
minimize rutting. This special provision is applicable to the wearing, binder and base
course mixtures used on the main line, ramps and cross-overs of all interstate highways ,
and other highways carrying more than 20,000 ADT or more than 1,000 daily Equivalent

l8 Kip Single Axle Load (ESAL) applications.

Table 2.4 :Summary of mix densities (lbs/c.ft), (31).

 

 

 

 

 

 

 

 

 

 

Asphalt Aggregate Top Size
Content
’/4" 1" 1 ' 2" 2" 2'/z"
3% - - - 148.77 150.00
4% 143.71 145.50 148.19 151.30 151.38
5% 144.14 148.02 149.68 150.25 149.65
6% 142.96 145.88 148.98 - -

 

 

 

 

Percent Voids in Mineral Aggregate

Percent Air Voids in the Mix

33

 

   

 

 

 

18 ,
l
16 .22 l iiiiii l
‘ Asphalt
“Lb l Content'/.
12 «~— ': ,1
1o ‘1 1 l +AC5%
“”7 T T i
1 1 ‘ +AC4%
84~ . . .
[ l f 1 +AC6%
l
5.2 , - 1 l i’ , w +AC3%
l l 3 l
44.? i i 7 i ‘ 7 l 7. 2 i 7 21‘, n 7 7
1 l l l
2‘—”* z; *"'*’f‘*“**l******ir**”*
I l l
O l l l i
o 0.5 2 2.5

1 1.5
Aggregate Top Size (inches)

Figure 2.10: Effect of aggregate top size upon the percent voids in mineral aggregate, (31).

 

Asphalt Content '/.
by Weight of
Aggregate
—e- AC 3%
+ AC 5%
+ AC 4%

+ AC 6%

 

 

 

 

o 0.5 1 1:5 2 2.5 3
Aggregate Top Size (inches)

Figure 2.1]: Effect of aggregate top size upon the percent air voids in the mix, (31).

34

The salient features contained in the special provision are:
- Use of A020 asphalt cement only;
- Use of larger size aggregate for binder and base course mixtures;
- More angular coarse aggregate. This is being increased to at least 85
percent two-face crushed particles;
- At least 75 percent manufactured sand in the fine aggregate;
- At least 4 percent minus No.200 fraction in the mixture to stiffen the
asphalt.
The binder type does not appear to be as important as the gradation in minimizing early
rutting. This conclusion is certainly related to the gradation used and possibly the type of

aggregate (13).

2.2.2.4 Sand and Mineral filler

The angularity and shape of the sand particles, the percent sand content in the AC
mix, and the type of mineral filler influence the rut and fatigue life performance of asphalt
pavements. The Federal Highway Administration (FHWA) recommends that natural sand
(rounded sand particles) be limited to 15 to 20 percent of the total weight of aggregates for
high volume roads and to 20 to 25 percent for medium and low volume roads (33).

Similarly, Button et a1. (34) suggested that for high traffic volume roads limiting the
natural (uncrushed) sand particle content of the asphalt mixes to about 10 to 15 percent
reduces rut potentials.

Young (35) cited Barksdale and Hicks, and reported a significant increase in

pavement rutting due to an increase in the percent sand contents. It appears that mixes with

35

less than 20 percent natural sand in the fine aggregate contained fewer fair to poor
pavements than mixes with more than 20 percent natural sand in the fine aggregate (32).

The significant effect of a reduction in fines content of only 3.0 percent from the
prescribed gradation is surprising. On the other hand, it is logical when one considers
that this essentially cuts in half the volume of fines in the mixture, reducing the amount of
cementing material in the mix, and changing the asphalt film thickness at the points of
aggregate contact. This indicates that variations of fines content, even within typical

gradation specifications, can have serious effects on mix performance (11).

2.2.2.5 Asphalt Type and Content

The type of the asphalt cement and its content in an asphalt mix impact the rut of the
asphalt mixes. Typically, the type of asphalt cement used in an asphalt mix depends on the
environmental conditions. Higher temperature regions require harder (higher viscosity or
lower penetration) asphalts. On the contrary, softer asphalt’s are used in the construction of

asphalt pavements in cold regions. The asphalt cement hardness or viscosity affects three

types of distress.
1. Fatigue life which is affected by the AC viscosity at in service temperatures.
2. Rutting- Softer asphalt tends to cause higher rutting potential. Hence rut is affected

by the AC viscosity at high temperatures.
3. Cold temperature cracking: Stiff asphalt cements tend to harden at lower
temperatures. Thus, cold temperature cracking is affected by the AC viscosity at

low temperatures.

36

The effects of the asphalt binder on the various distress modes vary. For example, a
study conducted by the National Research council, Strategic Highway Research Program
(SHRP), found that (4):

1. The binder contribution to the rut potential of the asphalt course is about 40 percent
as shown in Figure 2.12. The other 60 percent is attributed to the other ingredients
in the mix (aggregate, sand and mineral filler) and their proportions and to
construction practices.

2. The 40 percent is related to the pavement layer thickness and properties. Finally,

, the asphalt binder contribution to the low temperature cracking potential is about 80
percent.

Traditionally, asphalt binders are classified and specified according to their
viscosities at a given temperature. Two AC binders of the same grade (e.g., AC-20) may
have the same viscosity at the specified temperature. However, their viscosities at higher or
lower temperatures may differ substantially and so their behaviors. This fact has been
recognized in the new SHRP binder specifications which consider three types of distresses:
permanent deformation, thermal cracking and fatigue cracking . These distresses are related
to the rheological properties of the binder at high, low, and intermediate temperatures,
respectively (36).

Huber and Heiman (8) stated that penetration and viscosity of the AC binder do not

possess a significant effect on rutting rate because an asphalt grade typically produces a

Percent Contribution AC binder

37

 

100

 

 

Rutting Fatigue Low

Cracking Temp.
Cracking
Type of Distress

Figure 2.12: Binder contribution to difference distress (4).

 

38

similar penetration and viscosity after mixing. They attributed the rutting performance to air
voids, voids filled with asphalt and asphalt content.

Huges and Maupin (13) on the other hand, stated that binder type is not as important
as aggregate gradation in minimizing early pavement rutting.

Kruts (37) evaluated the impact of moisture on rutting potential of polymer
modified mixtures. He reported that moisture conditioning of laboratory specimens
yielded fairly consistent permanent deformation results for all mixtures regardless of the
grade of the asphalt cement binder. An excessive amount of asphalt or a very soft grade
asphalt can often produce a mix with high Marshall flow values. A maximum Marshall
flow of 16 is often specified for mix design and construction control. Brown and Cross (38)
reported that Marshall flow is a good indicator of rutting potential. Higher amount of
rutting is associated with mixes having Marshall flow values above 10.

Mixes with large aggregates have less asphalt content requirement to coat the
aggregate. Case studies of rutted pavements in Oklahoma conducted by Hensley and Leahy
(39) showed that missing intermediate fines in stone-filled mixtures make them less
sensitive to asphalt content. This is due to the fact that intermediate fines increase the
surface area to be coated with asphalt cement and increase their ball bearings effects. Thus,
their absence not only reduces the required asphalt content but also reduces the sensitivity
of the mix to increased asphalt contents.

Decker and Goodrich (40) ranked five asphalt cement factors relative to their effects
on asphalt pavement distresses. They used a ranking scale from O to 5. A ranking of 0
indicates no effect whereas 5 indicates significant effect. Since, asphalt cement properties

are mainly governed by the source of the crude, they found that the refining process has a

39
very minor effect on rutting and fatigue cracking. Decker and Goodrich studied and ranked

the physical properties (such as stiffness, temperature susceptibility) and rheology of the
AC binders (see Table 2.5). They concluded that increase in the viscosity of the AC binder
causes an increase in the stiffness of the AC mix which in turn improves the rut resistance.

For asphalt pavements with high rutting in the top courses of the asphalt concrete,
it was observed that these courses often contained fine aggregate gradations and high
asphalt contents. One of the biggest causes of rutting is excessive asphalt content in
asphalt mixtures. Steps should be taken to ensure that the proper asphalt content is
selected and provided during mix production. The asphalt content should be selected to
provide void contents of 4 to 5 percent in laboratory compacted mixtures for high traffic
volume roads. Asphalt content should not arbitrarily be increased to facilitate
compaction, to minimize segregation, or for any other reason except to provide
satisfactory voids in the laboratory compacted asphalt mixture (3 8).

The study in Alabama revealed that the asphalt content is also critical as the mix
reaches about 4% voids structures. Excessive asphalt will decrease inter granular
contacts, weakening the aggregate skeletal structure and leading to further densification.
Excess asphalt can weaken otherwise very stable aggregate structures (10).

According to Barksdale, rutting was found to be directly related to asphalt
content, it was not extremely sensitive to material type, gradation and level of compaction
used in the Marshal Mix Design Method. For the relative low asphalt content of 4.5 to 4.8
percent presently used in base course mixes in Geogria,. An increase 1/4 to 1/2 percent

asphalt should result in rut depths of about 0.20 to 0.25-in (18).

40

Table 2.5: Effect of asphalt cement factors on pavement distress (40)

 

 

 

 

 

 

 

 

PAVEMENT PERFORMANCE
ASPHALT FACTORS THERMAL FATIGUE RUTTING MOISTURE
CRACKING CRACKING DAMAGE

REFINING TECHNIQUE 1 l 1 0
PHYSICAL PROPERTIES 4 3 2 1.5
CHEMICAL PROPERTIES 2 2 2 2
AGING 1 3 1 1.5
MODIFICATION 2.5 2.5 2.5 2.5

 

 

 

 

 

 

41
Based on the observations of Dhahran - Abqaiq Road in Suadi Arabia, it was

concluded that one of the reasons for rutting at the Abqaiq section was overasphalting
(4.61% asphalt content in the field) which also indicates that the Marshal stability test
might not be appropriate for the measurement of properties of asphalt mixtures,
especially for the crushed limestone. Some of the potential reasons for the rutting were

over asphalting, fine-graded mix, and poor quality control (12).

2.2.2.6 Environmental Factors

Exposure to environments causes the bituminous material to harden over time. With
passage of time the bituminous binder becomes so brittle that it can no longer sustain the
strains, affiliated with daily temperature changes and traffic loads. The rate of hardening is a
fimction of oxidation resistance of binder, the air voids content, temperature, and thickness
of the asphalt fihn. Therefore, the rate of hardening varies with the binder type, climate and
material design. It should be noted that most of the asphalt hardening takes place during
mixing, agitating, transporting, and construction.

Asphalt durability can be defined as its resistance to change in its original properties
during mixing, construction, and service life. Durability of the asphalt is not a factor that
effects rutting directly. In fact the embrittlement of the asphalt due to aging can cause
increase in both the cohesion and the stiffness of the binder, which result in a greater
resistance to permanent deformation under loading. Asphalt hardening on the other hand,
adversely affects its fatigue life and low temperature cracking. Further, aging of the binder
is not sufficient to deter rutting in an under designed pavement structure. Cases have

been reported where rutting has occurred suddenly in older pavements which were

exposed to higher tire pressure and heavier axle loading than expected (3 8).

42
Huges and Maupin (10) reported that early pavement rutting is a function of the

temperature of the pavements when it is opened to traffic. They suggested pavement

temperatures of leSs than 150 °F lead to a stable asphalt mix under traffic.

2.2.2.7 Ranking of Factors
A comprehensive study of the effects of asphalt mixture variables on the

performance of flexible pavements and on the engineering properties of the mixtures was

sponsored by the National Cooperative Highway Research Program (N CHRP). The title of

the project is Asphalt Aggregate Mixture Analysis System (AAMAS). The asphalt mix

design factors included in the study are: compaction, resilient and creep modulus, indirect

tensile strength and strain at failure and the compressive strength. Each factor was ranked
on a scale from 0 to 5. A ranking of 0 indicates that the factor has no effect on the pavement

performance. A ranking of 5 indicates significant effects (11).

Two static creep and repeated load tests were used in the AAMAS study, the
uniaxial compression and the indirect tensile. The values of the resilient modulus obtained
from the test were used to evaluate the rut potential of the compacted asphalt mixes.

Table 2.6 provides a list of the ranking of the various test outputs. It can be
seen that:

1. The creep modulus of the AC mixture has significant effect on the rut and thermal
cracking potentials, a moderate effect on fatigue cracking, and a minor effect on
moisture damage potential.

2. The compressive strength has a moderate effect on rut potential, a minor effect on
moisture damage, and no effects on the fatigue and thermal cracking potentials of
the mixes.

Other mix factors such as aggregate particles orientation, which is affected by the

compaction method (the gyratory compaction technique was recommended by the AAMAS

43
study), the slope and intercept of the creep-time curve (also known as the alpha and gnu

parameters), the tensile stress to tensile strength ratio, and the compressive stress to
compressive strength ratio were also included in the AAMAS study. Their respective
ratings are listed in Table 2.6.

Polymer modifications resulted in improved rutting resistance in wheel tracking
tests, carried out by Collins, Bouldin and Berker. They reported that, with increasing
polymer content, the systems will be more resistant to deformation and exhibit enhanced
elastic recoil. For this specific mix valkerting, et al found that already at polymer
concentrations as low as 2 % by wt, the rutting resistance was increased by a factor of
approximately 2. The blend containing 6 % by wt leads even to a 7.5 fold increase (41).
2.2.2.8 Rut Prediction Models

The proper pavement design and pavement management systems must be capable of
predicting the performance, and the remaining service life of the pavement structures. Rut
is one type of distress that must be predicted during the pavement design and the pavement
evaluation processes. In several countries, an overlay is applied when the rut depth of the
pavement is of the order of 20 to 30 mm. In the USA, the maximum acceptable rut varies
from one State Highway Agency (SHA) to another and it depends on the pavement class
(interstate versus farm to market roads). Current rut prediction models can be divided, in
general, into two groups: mechanistic and mechanistic/empirical. The mechanistic models
are based either on the theory of elasticity, theory of plasticity, or the viscoelastic theory.

Barksdale (42, 43), Heukelon and Klomp (44), Romain (45) and others have
suggested that layered elastic theory can be used to calculate the induced stresses and strains
in the pavement due to traffic. The plastic strain can be assumed to be proportional to the

elastic strain and the number of load repetitions.

44

Table 2.6 : Summary of Ranking of Mix Properties Related to Pavement

 

 

Performance (11).
FACTORS DISTRESS MANIFESTATIONS
Fundamental
Engineering Permanent Thermal Fatigue Moisture

Properties Deformation Cracking Cracking Damage

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

Resilient Modulus 3 3 5 1
Creep Modulus 5 5 3 l
TensfleTtrarn at

Failure 1 4 5 1
Indirect Tensrle
Strength 3 0 0 1
Other Properties & Factors
Particle Orientation 3 3 3 0
Alpha and Gnu 5 1 2 0
Tensile Strength Ratio 2 0 3 5
‘ Resrlrent Modulus
Ratio 3 0 2 4
'Com—p' r'essrve Sfi‘engtn
Ratio 2 0 O 2

 

 

 

 

 

 

45
In this type of analysis, the relationship between the plastic strain and the applied stress for

each pavement component can be obtained from laboratory repeated load test along with
either linear or nonlinear elastic theory. Nonlinear theory should provide more accurate
results, but its use has been limited because of its complex nature. Rut prediction cannot be
made directly from plastic stress-strain relationship although, it provides a considerable
insight regarding material behavior.

Barksdale and Leonards (46) suggested that rutting can be estimated by assuming that
the pavement can be represented as a viscoelastic layered system. The plastic strain can
then be predicted by using the material characteristics obtained in the laboratory from the
creep tests. Elliot and Moavenzadeh (47) suggested the use of the linear rather than the
nonlinear viscoelastic theory due to its mathematical complexities for predicting structural
response.

Morris (3) cited Kenis and a study conducted at Washington State University Test
Track and reported that the rut prediction obtained by using linear viscoelastic theory were
substantially lower than the actual measured rut depth. For the test track, after 247,000 load
repetitions, the linear viscoelastic theory predicted rut depth in the order of 5 to 10 percent
of the actual measured ones.

The complexities involved with the theoretical prediction models tempted the
researchers to develop mechanistic/empirical models to predict rutting. Some of these
models are presented below.

Leahy and Witczak (48) assessed the influence of repeated triaxial test conditions
and mix parameters on the permanent deformation characteristics of asphalt concrete and
presented a statistical model of the form:

logwep: 15.83+7.132[log,0('1)]+1 .105[log.0(S)]0.1 18[log,o(V)]+2.155[log,O(EAC)]

+ 1.117[loglo (VOL)]+[0.986(T1'EMP‘0''02 VMA‘O'”8 )loglo (LN)] (2.1)

Where;

'0

<m—1

EAC
VOL
TTEMP
VMA
LN

46
R2 =0.82

= permanent deformation;

= test temperature (°F);

= deviator stress (lb/inz);

= viscosity at 70° F (106 poise);

= effective asphalt content (percent by volume);
= percent air void;

= test temperature (°F);

percent void in mineral aggregates; and

= load repetition.

A sensitivity analysis of the above equation showed that the temperature is the most

significant variable. The model is less sensitive to loading conditions, material type and mix

parameters(48).

Based on laboratory test data Morris et a1. (9), presented the following regression

equation, to predict pavement rutting:

Where

sp=f (SI, 52 ,T,N) i E (2.2)

amount of permanent strain;

vertical stress;

horizontal stress;

temperature;

number of load applications; and

The estimate of error associated with any attempt to predict

8p as a function of the other four factors.

47

The laboratory developed model was used to predict the observed rutting at the

Barmpton Road Test. Although the model does not account for the effect of the AC

thickness, the binder type, and the percent air voids, the overall predicted and observed

values were comparable as shown in Figure 2.13.

Baladi (49) correlated AC mix and other pavement layers properties to predict the

rut depth in the AC layer. He presented an equation of the form:

LOG(RD) =

where

-1.6 + 0.067(AV) - 1.4[1og(TAC)]+ 0.07(AAT)
- 0.000434(KV) + 0.15[log(ESAL)] - O.4[log(MRRB) ]
- 0.50[log(MRB)] + 0.1[log(SD)] + 0.01[log(CS)]

- 0.7[log(TBEQ)] + 0.09[log{50-(TAC + TBEQ )}] (2.3)

LOG natural log;

RD rut depth (inch);

AV the percent air void in the mix;

TAC thickness of AC layer (inch);

AAT average annual temperature (°F);

KV kinematic viscosity at 275 °F (AASHTO T-201)
(centistrokes);

ESAL the number of 18-Kip ESALs at which the rut depth is being
calculated;

MRRB resilient modulus of roadbed soil (psi);

MRB resilient modulus of the base material (psi);

SD pavement surface deflection (inch);

CS compressive strain at the bottom of the AC layer; and

48

 

 

 

 

to
3' 0.8-
". -—o-- amen
g ...—X_ "man
:
g onl- I ssss cosmetic: sum
c e
o
u.
3 . ‘ I’e
.- O"- r
E I /’
i A.
3 oz- I ‘1’!
’
I
L l l l l

to p. C O O - a a

2 2 2 2 '2 2 2 ~

i i 8 2 8

5 -s g g g g 3

TIME

Figure 2.13 : Permanent Deformation as a Function of Time (Brampton Road Test
Section 3) (9)

49
TBEQ = equivalent thickness of base material (inch), which is the

actual thickness of the base layer plus the equivalent
thickness of the subbase layer; equivalent thickness of the
subbase layer is equal to the actual thickness of the subbase
layer reduced by the ratio of the modulus of the subbase to
that of the base material.

Baladi, concluded that air voids is the most important factor affecting rut and higher
air voids lead to higher rut potential.

Based on a study of 19 pavement sites conducted at Michigan State University,
Mukhtar (4) developed an empirical rut model to predict pavement rutting. The model is of

following form :

RUT = ESAL [-3656 + .00068 {(FINE)3 +SAND} +.00012 {EXP(AV)}

+ .0289 (sz) - .0037 (AV) + .00314 (AC3) -.6614 (LOG(ANG))] (2.4)

where

Rut = rut depth (inch);

ESAL = 18-kip cumulative ESAL during the service life;
AC = asphalt content (percent by weight);

SAND = sand in AC mix (percent by weight);

FINE = fine in AC mix (percent by weight); and

50
AN G = angularity of coarse aggregate (plus #4) in AC mix.

Basing on the sensitivity analysis of the model Mukhtar concluded :

1. There is no appreciable increase in rut depth as the percent fine content increases
from 0 to about 5 percent. Further increase in the percent fine content causes
substantial increase in pavement rutting.

2. Increasing the percent air voids from 3 to 6 percent causes an insignificant increase
in the rut depth. Increasing the percent air voids above the 6 percent level causes a
substantial increase in rut.

3. Increasing asphalt content fi'om 4 to 8 percent causes an increase in the rut depth.

4. Increasing aggregate angularity fiom 1 (rounded aggregate such as river gravel) to 5

(aggregate crushed on all sides) causes a substantial decrease in the rut depth.

Miller and Ford (20) carried out regression analysis to determine the most
significant relationships between pavement ruts and mixture of physical measurements.
The pavement mixture characteristics found to have an appreciable coefficient of
correlation with rutting or the logarithm of rutting include, respectively, air voids, 0.674;
voids filled, 0.658; and resilient modulus, 0.602. The effect of the amount and type of
asphalt and the aggregate character may be reflected in the above factors. Stepwise linear
regression analysis was conducted to determine the best-fitted equation for each
dependent variable and its relationship to other mix characteristics. One best-fitted
equation that illustrates the relationship between rutting and the mixture properties is the

following:

RUT = - 73.3 + 0.937 (VF) + 0.582 (D40) + 2.33 (BAV) - 0.0236 (STAB) (2.5)

51

where

RUT = rut depth, 1/32 in.;

VF = voids filled (percent);

D40 = hump in grading curve at #40 sieve (percent). The value is determined on

a plot of the gradation on 0.45 power paper. A line is extended from
origin to #4 sieve data point. The difference in % between the straight
line and the gradation at #40 is hump;

BAV = air voids between wheel path (percent);

STAB = Marshal stability (psi).

The coefficient of determination (R2) value is 0.495 for this equation with a high
confidence level. This indicates that only about 50 percent of the rutting is explained by
this equation. Additional factors to be considered in design that affect rutting include
traffic speed and character, environmental conditions, longitudinal grade and support
from the underlying pavement structure and subgrade. These factors were not part of this
analysis. Their evaluation would increase the understanding of the rutting problem. The
results of this study indicated that the mixture air voids of 2.5 to 5 percent will provide
asphalt mixtures that have an acceptable level of rutting. Deep ruts are associated with
pavements having air voids of less than 1.0 percent and with the traffic that is slow
moving and channeled (20).

Thompson and Naurnan (50) developed a model to predict permanent strain as a
function of load repetitions. The model is expressed as:

Log E, = a + b Log N (2.6)

or: E? = ANb

where: E? = permanent strain;

a & b = experimentally determined factors; and

A = antilog of “a”.

52
Ohio State University (OSU) researchers have proposed a permanent strain

accumulation prediction model for use in a pavement design system developed for the
ohio DOT. The OSU permanent strain accumulation model is:

EP/N = Anm (2.7)

where

5,, = plastic strain at N number of load repetitions;

N = number of repeated load applications;

A = experimental constant dependent on material and

state of stress conditions; and

m = experimental constant depending on material type;
Note: If the “b” term from the log strain - log N model is

known, m is equal to b-l (50).

James Mark Matthews and BB. Pandey, (51) developed a regression equation
basing on the performance of flexible pavements:

RD = -0.265 + 6.79 ( SVS )+ 3.08 (ESAL) (2.8)

where

RD = average rut depth (m),

SVS = subgrade vertical strain (x 10-3) obtained from

the computer output (RAOPAVE), and

ESAL = (x 107).

Equation 2.8 has a squared multiple regression coefficient of 0.951, a calculated F
value of 116.0, and a critical F value 6.93 at 0.01 significance level. Because the

calculated F value is greater than the critical F value, the null hypothesis is rejected,

53

leaving the linear terms in the model. It would be useful, and seems a natural extension,
to relate rut depth to gravel surfacing depth and cumulative ESALs. This relationship is
shown by the following equation:

RD = 3.878 (ESAL) - 0.0459 (TWBM) (2.9)

where

TWBM equals the thickness of WBM, in millimeters.

Equation 2.9 has a squared multiple regression coefficient of 0.727, a calculated F
value of 32.0, and a critical P value of 6.93 at 0.01 significance level.

For granular pavements, the loss of serviceability is caused mostly by excessive
permanent deformation along the wheel-path. Hence, in this study rutting was adopted as
the index for judging pavement performance, which is in agreement with British practice.
An average rut depth of 15 mm is considered adequate for an overlay done at the optimal
time. Putting this value of rut depth in Equation 2.8, the following rutting model was
developed for the traffic range covered in this study:

SVS = 4.93 "‘ 105 (ESAL "'22) (2.10)

where SVS equals the subgrade vertical strain in 0.001 mm/mm.

Equation 2.10 was compared with that of Shell

SVS = 2.8 * 10‘2 (ESAL '0'”) (2.11)

The difference in the coefficients of these equations is caused by the difference in
the material characteristics. Although the difference of the corresponding coefficients in
the two equation is considerable, the curves are close to one another when plotted. For an
ESAL value of 30 million (a typical value - the ESAL range in this study is 8 to 45

million), both equations yield a subgrade vertical strain of 0.00038 mm/mm. An average

54

rut depth of 20 mm is considered adequate for the complete failure of a pavement
structure. For this rut depth, both equations gave the same subgrade vertical strain

(0.000348 mm/mm), for an ESAL value of 42 million (51).

CHAPTER 3

RESEARCH PLAN

3.1 GENERAL

Over the last two decades, the number of heavy multi-axle trailers in Pakistan has
risen many fold. This resulted in premature manifestation of rutting and its rapid
development of high severity levels.

Pakistan today, has a prematurely and severely deteriorated road network of AC
pavements. These present loss of precious infi'a-structure worth billions of rupees. If this
problem (rutting) is continued to be neglected, then new AC pavement projects will also
crumble prematurely. The rehabilitation costs will be a formidable obstacle to the socio-
economic development of the country.

The resistance of the AC mixes to rutting is a function of the AC mix properties
(percent coarse and fine aggregate, angularity of the aggregate, and the asphalt binder
content and viscosity, air voids ), the pavement design process, construction practices and
environmental factors. Recall that (see Chapter 2) one of the SHRP studies concluded that
forty percent of rutting, sixty percent of fatigue cracking and eighty percent of thermal
cracking potential may be attributable to the asphalt binder alone whereas, the remaining
percentages of the respective distresses can be attributed to the other properties of the AC
mix and to the overall design of the pavements.

Analysis of data from field test sites in Alabama indicates that permanent

deformation causing rutting is generally confined to the top 3 to 4 inches (surface and
55

56

binder courses). There was little evidence that the lower base/subbase layers or the

subgrade were significant contributors to rutting (10). This suggests that rut resistance of

AC mixes can be enhanced by changing the properties of the AC mixes to withstand the

present day traffic conditions and to perform satisfactorily over the design life. Research

studies need to be conducted to establish the AC mix factors that enhance the resistance

against rutting of the mix.

3.2

RESEARCH PLAN

To accomplish the objectives of this study (see chapter 1 ), a research plan was

framed. The plan comprises of seven steps as follows(figure 3.1 ):

1.

Select pavement sections having various asphalt grades and percent asphalt
contents, aggregate types and angularity, gradation, various types of fines and
percent fine contents, various AC thicknesses, different levels of traffic volume and
load, and different service lives.

Obtain traffic data (traffic counts and axle loads) for each selected section.

Secure construction history (date of completion/date of opening traffic on the road
sections) from the National Highway Authority (NHA) of Pakistan and other
consultant/construction agencies.

Measure the rut depths at different site/sections, the longitudinal grade and obtain
core samples.

Conduct laboratory tests on the core samples to determine the physical(AC layer
thickness) and asphalt mix properties. These include aggregate gradation, asphalt
content, bulk specific gravity, penetration of recovered asphalt, and the angularity

of the aggregate.

 

 

 

Traffic
Data

 

 

57

 

 

 

 

 

Construction
History

 

 

 

 

 

 

 

Asphalt Mix Properties
from Cores & Lab Tests

 

 

 

Statistical
Analysis -

 

 

 

 

 

Rut

 

Determine
Factors
Affecting
Rutting in
Pakistan

 

 

Conduct
Sensitivity
Analysis.

 

 

 

 

 

Measurements

 

 

 

Conclusions
&
Recommendations

 

 

 

Figure 3.1: Overall Research Plan

 

 

58

6. Perform statistical analysis relating the rut depths to the AC mix properties, traffic
and road attributes.
7. Conduct sensitivity analysis whereby the effects of the various parameters of the

statistical models on the rut life are assessed.
3.2.1 Pavement Selection and Investigations
The selection of pavement test section/site/locations was accomplished in
consultation with personnel from the National Highway Authority (NHA) of Pakistan. The
subject study has been designed to analyze pavement sections in areas from Peshawar to
Sahiwal. In this regard, a fill] factorial matrix with the following two independent variables
has been formed. (see Figure 1.3)
1. Asphalt Course Thickness - Thin (5 to 10 cm), moderate (10 to 18 cm), thick
(more than 18 cm) asphalt surface.
2 Traffic Volume and loads - Light (0 to 2), moderate (2 to 4), heavy (>4) million 18
kips equivalent single axle loads (ESALs).
The matrix consists of 9 cells, each cell represents pavement sections, to be selected.
Each section will have variable length of 1 to 5 kilometers depending on the number of
sites, the various degrees of severity levels of rutting (See Figure 1.3). Each test site is 30
meter long and it contains 5 equally spaced test locations (See Figure 1.4).

For each test site/location, the rut depth was measured and one core sample was
obtained. Because of economical reasons and time constraints, the number of cores was
limited (one per each site).

The rut depths were measured by using a 3 & 2 meter straight-edge leveling rods

with a graduated steel ruler with an accuracy of l to 2 mm. The ruts were measured in the

59

outer wheel path over each marked test location (five test locations 6 meter long at each test
site). The average of the five measurements was then calculated for each test site.

The cores were brought to the laboratory, where the sample thicknesses were
recorded and tests were performed according to the American society for Testing &
Material (ASTMs) Standard test procedures. The laboratory tests were designed to
determine, the bulk specific gravity, gradation, the asphalt content, recovered penetration of

the asphalt and the angularity of the aggregate ( for more details, see chapter 4).

4.1

CHAPTER 4

FIELD AND LABORATORY INVESTIGATION

FIELD INVESTIGATIONS

4.1.1 Pavement Selection

The pavement section selection was accomplished by surveying the routes

physically. The main criterion used in this selection is that the selected sections should

represent the spectrum of pavement cross-sections, paving materials, and traffic volume

and load found in most areas in Pakistan. In this regard, the following variables were

identified and prioritized prior to the selection of the pavement sections.

1.

Asphalt Course Thicknesses - Thin (5 to 10 cm), moderately thick (10 to 18
cm), and thick (more than 18 cm) asphalt surface.

Traffic Volume and Load - Heavy, moderate, and light traffic loads and volumes
in terms of the 18-kip equivalent single axle load (ESAL).

Pavement Type - flexible pavement without overlays.

Cross-Sections - Three layers (AC, base and subbase).

AC Mixes - Stability based and standard type mixes.

Roadbed Type - Cohesive soils.

Pavement Surface Age - Newly constructed(less than 2-years old) and older
pavement sections.

Distress Type - Rutting.

Based on the above criteria and the various variables, 54 flexible pavement

sections of variable lengths (one to five kilometers) were initially selected. For each

60

61
pavement section, the National Highway Authority (N HA) referencing system (kilometer

stone, distances measured from Karachi) were used to obtain the location reference point.
The construction and rehabilitation history, was obtained from NHA and different
consultant/construction agencies in Pakistan. The rut depths and the longitudinal grade
were measured. Certain sections had to be excluded from the study due to lack of data
regarding date of completion/opening traffic of these sections. Traffic volume and load,
were acquired from the National Transport Research Center (NTRC) Islamabad, a traffic
study conducted by Mr. Zafar Iqbal Farroqui, and other Provincial Highway agencies.
Rut depths, cross-sectional data (layer thicknesses and types), were obtained by actual
field/laboratory measurements. The data was then tabulated in a spreadsheet. Table 4.1
provides lists of the flexible pavement sections chosen in this study. Other data such as
pavement type, route number, direction (north, south, east, or west bound), section
number, and the beginning and ending kilometer post of each pavement section are also
listed in the Table.

Each pavement section was examined/surveyed and, within each section, several
100-feet long pavement sites showing rutting problem were identified and they were
labeled as test sites. For some pavement sections, the test sites were adjacent to each
other while for some others, they were quite apart. During the Winter of 1994, a four
member research team of the Military College of Engineering (MCE) visited each test
site. The purpose of the visit was to:

1. Verify the location reference point, pavement type, and general conditions.

62

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

Table 4.1 : Selected Flexible Sections
Section Route Direction Kilometer Post Distance Distance Sample
& NH = 11 N = 01 From Karachi From From Designation
SITE PH = 22 E = 02 Rawal-pindi Main Number

LH = 33 S = 03 From To Road

W = 04

3A 11 4 1696 1697 156 - 11043A156
4B 1 1 4 1685 1699 145 - 1 10448146
4C 11 4 1685 1699 159 - 11044C159
5A 11 4 1669 1670 130 - 11045Al30
58 ll 4 1669 1670 130 - 11045C130
6A 11 4 1661 1662 121 - 11046A121
6B 11 4 1661 1662 122 - 110468122
7B 11 4 1656 1657 116 - 110478116
7C 11 4 1656 1657 116 - 11047C116
88 11 4 1641 1642 102 - 110488102
9A 1 l 4 1637 1638 98 - l 1049A98
9B 1 1 4 1637 1638 98 - 11049898
10A 11 4 1637 1638 97 - 110410A97
108 11 4 1637 1638 97 - 110410897
12A 11 1 1577 1585 37 - 110112A37
12C 11 l 1577 1585 44 - 110112C44
13A 11 l 1555 1556 15 - 110113A15
138 ll 1 1555 1556 15 - 110113815
14A 11 1 1561 1564 22 - 110114A22
148 11 l 1561 1564 22 - 110114822
14C 11 1 1561 1564 24 - 110114C24
16A 33 2 1550 1550 10 l 330216A101
17A 33 2 1550 1550 10 8 330217A108
178 33 2 1550 1550 10 9 3302178109
18A 33 2 1540 1540 0 7 330218A07
188 33 2 1540 1540 0 8 330218808
19C 33 2 1540 1540 0 14 330219C014
208 11 3 1505 1506 35 - 110320835
21A 11 3 1508 1509 32 - 110321A32
22A 1 l 3 1497 1498 43 - l 10322A43

 

 

 

 

 

 

 

 

 

63

Table 4.1 : Selected Flexible Sections (Continued)

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

Section Route Direction Kilometer Post Distance Distance Sample
& NH = 11 N = 01 From Karachi From From Designation
SITE PH = 22 E = 02 Rawal-pindi Main Number

LH = 33 S = 03 From To Road

W = 04

238 l 1 3 1497 1498 42 - 1 10323842
24A 1 1 3 1486 1487 53 - 1 10324A53
24B 1 l 3 1486 1487 54 - 110324854
25A 11 3 1478 1487 53 - 110325A53
268 11 3 1452 1452 88 - 1 10326888
27A 11 3 1452 1458 82 - 110327A82
27B 1 l 3 1452 1458 88 - 1 10327888
28A 1 1 3 1441 1442 98 - 1 10328A98
29A 11 3 1442 1442 98 - 110329A98
30A 11 3 1416 1417 124 - 110330A124
318 11 3 1416 1417 123 - 1103318123
31C 11 3 1416 1417. 124 — 110331C124
328 11 3 1408 1409 132 - 1103328132
33A 11 3 1408 1409 131 - 110333Al31
34A 11 3 1387 1395 145 - 110334A145
34B 11 3 1387 1395 153 - 1103348153
36A 11 3 1365 1366 175 - 110336A175
37A 11 3 1381 1382 159 - 110337A159
38A 11 3 1364 1365 176 - 110338A176
388 11 3 1364 1365 175 - 1103388175
42A 1 1 3 1291 1291 249 - l 10342A249
43A 1 1 3 1275 1276 265 - 1 10343A265
44A 11 3 1226 1227 314 - 1 10344A3 14
45A 11 3 1227 1228 312 - 110345A312
458 ll 3 1227 1228 312 - 1103458312
45C 11 3 1227 1228 313 - 110345C3 13
46A 11 3 1175 1179 361 - 110346A36l
468 ll 3 1175 1179 364 - 1 103468364
478 11 3 1175 1176 365 - 1103478365
47C 11 3 1175 1 176 364 - l 10347C364
48A 11 3 1125 1126 414 - 110348A414
488 11 3 1125 1126 415 - 1103488415

 

 

 

 

 

 

 

 

 

64

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

Table 4.1 : Selected flexible sections (Continued)
Section Route Direction Kilometer Post Distance Distance Sample
& NH = 11 N = 01 From Karachi From From Designation
SITE PH = 22 E = 02 RawaI-pindi Main Number

LH = 33 S = 03 From To Road

W = 04

49A 11 3 1122 1123 418 - 110349A418
51A 22 3 1250 1250 290 5 22035 1 A2905
52A 22 3 1268 1268 272 15 2203 52A272 l 5
528 22 3 1268 1268 272 25 220352827225
53A 22 3 1334 1334 206 l 220353A2061
54A 22 4 1661 1661 121 3 220454A1213
55A 11 1 1585 1585 45 3 110155A453
58A 11 4 1554 1559 14 - 110458A14
58C 11 4 1554 1559 19 - 110458Cl9
59A 1 1 4 1586 1587 46 - 1 10459A46
598 1 l 4 1586 1587 47 - 1 10459847
61A 33 2 1540 1540 0 14 330261A014

 

 

 

 

 

 

 

 

 

65

2. Mark the test sites.
3. Inspect, measure, and record the extent and severity of rutting.
4. Identify those test sites to be cored.

4.1.2 Measurements of Rut Depths

The rut depths were measured by using a three/two meter straight-edge leveling
rod, a graduated steel ruler with an accuracy of 1 to 2 mm. The ruts were measured in the
outer wheel path over each marked core location and at five other test locations along the
30 meter long test site and the average rut depth was calculated. For each test section of
1 to 5 kilometer length one or more test sites were selected, depending upon the
availability of severity levels of rutting. For each severity level, one test site was

identified (For more details refer to Table 4.3).

4.1.3 Core Locations

A total of 75 locations were designated for pavement coring. Cores were taken for
three severity levels within one test section depending upon availability. Core locations
were given specific designation numbers. The coring method and the designation

numbers are presented in the next section.

66

4.1.4 Marking Coding and Caring

Each test site was designated by a number and an alphabet system . The number
designates the pavement section and the alphabet designates the test site. For example, a
test site designation of 29-A indicates the first test site of pavement section number 29.

All test sections were cored. The cores were extracted from the outer wheel path,
at the start of the test site. Each core location was designated using a nine digit number
and a letter. Starting at the left-most digit, the first two digits indicate the type of route
(National-11, Provincial-22 or local-33), the next two digits designate direction from base
station, Islamabad,(North-01, East-02, West-03 and South-04), the fifth and sixth digits
indicate the section number, the alphabet designates the site number, the eighth digit
indicates the distance of the core location from Islamabad. The ninth digit indicates the
distance of core location from a main route(if located on the branch route). For example,
a core location designation of 11043A156 indicates (left to right) that the test is
conducted on section of National Highway, west of Islamabad, test section number 3 and
site A, located 156 kilometers from Islamabad.

The cores were obtained by using a powered auger equipped with a 6/4-inch
coring bit. For thin AC layers 6 inch diameter coring bit was used and for thicker layers 4
inch coring bit was used. These produced enough material for testing according to
ASTM. Most of the cores obtained from the wheel paths were utilized for measurements
of bulk specific gravity, layer thickness and extraction testing to determine material

properties. The laboratory test procedures are presented in section 4.3 of this chapter.

67

4.1.5 Traffic Volume and Loads

In view of the calculation of yearly and cumulative ESALs over the designated
sections, the traffic volumes (the number of trucks of different configuration, passing
over a test section in 24 hours) were obtained from the records of National Transport
Research Center in Islamabad (Traffic Volume Data 1994-95), Provincial Governments
and the traffic counts of Capital Development Authority in Islamabad traffic (52, 53, 54).

The annual growth rates of trucks calculated by NTRC from 1983-93 were used
(55). These rates are shown below. In this study, a traffic growth rate of 8 to 9 percent per
year was used.

Years Annual Growth Rate

1983-87 6%
1987-90 7.7%
1990-93 9.8%

To calculate the ESALS, different axle load studies carried out in Pakistan, were
consulted. These include the axle Load survey 1993 (Taxilla to Muredike) by Zafar Iqbal
Farooqui (56), Axle Load Study on National Highways 1995 carried out by NTRC (57),
and Axle Load Studies in Punjab (58). Out of these studies, Mr. Farooqui and NTRC
studies were found relatively more accurate and they were used in this study.

Truck factors of different categories of truck in Pakistan over the designated

pavement section were used. In case where the Truck Factor of a particular section was

68
not available, the nearby Truck Factor was used. AASHTO LEFs (Pt=2.5 and SN=5) and

growth factor has been used in calculation of truck factor and cumulative ESALS.
Some of the truck factors at a particular section are appended in Table 4.2 for reference
(56,57,58)
4.1.5.1 Definition - Truck Factor

Number of ESALS applied by each vehicle is known as ESALS factor. Since the
majority of the load on a pavement is due to truck traffic. The vehicle specific ESAL
factor commonly called a truck factor which is in unit of ESALS/T ruck, is most
important factor in computing the design ESALS (Commulative ESALS) (56, 57, 58).
4.1.5.2 Service Life

The ages of test sections and test lsites were calculated based on the date of
opening the pavement to traffic (obtained from records of NHA and other
consultant/construction agencies) and the date of taking rut measurements. The data is

shown in Table 4.3.

4.2 LABORATORY INVESTIGATIONS

This section provides a summary of the laboratory test procedures used to
determine the physical and engineering characteristics of the materials. Various
laboratory tests were conducted to assess the properties of the core materials obtained
from the various test sites. Since these tests are standards of the American Society for

Testing and Material (ASTM), only the ASTM test designation numbers and the purpose

69

Table 4.2: Truck Factors at Taxilla (56, 57)

 

 

 

 

 

 

 

 

 

 

 

Axle Configuration Truck Factor
1.2 6.03

1.22 1 1.851

1.2+22 4.386

1.2+2.2 6.996

1.2+222 14.73

1.22+222 15.821

1.1+1 2.264

1.1+2 3.35

 

 

7O

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

Table 4.3: Field Data - All Test Sites
Section & Sample Sample Yearly Service Cum Rut
Site Designation Thick ESALS Life ESALS Depth
Number (cm) (millions) (years) (millions) (cm)
3A 1 1043A156 8.9 0.850 8.000 4.250 0.953
48 110448146 8.9 1.250 8.000 6.250 0.953
4C 1 1044C159 9.8 1.250 8.000 6.250 2.096
5A 11045A130 11.4 1.000 8.000 5.000 0.635
58 1 1045C130 10.2 1.000 8.000 5.000 1.270
6A 1 1046A121 8.9 1.250 8.000 6.250 1.143
68 1 10468122 10.2 1.250 8.000 6.250 1.905
78 110478116 5.1 2.023 7.435 9.613 1.270
7C 11047C116 5.1 2.023 7.435 9.613 2.54
88 110488102 5.1 2.800 7.435 13.340 2.54
9A 1 1049A98 10.8 2.030 7.435 9.680 0.953
98 1 1049898 16.8 2.030 7.435 9.680 2.223
10A 110410A97 12.7 2.800 7.435 13.340 1.270
108 110410897 1 1.4 2.800 7.435 13.340 2.223
12A 110112A37 11.4 3.070 1.403 5.310 0.635
12C 110112C44 7.6 3.070 1.403 5.310 2.159
13A 110113A15 15.2 10.810 0.890 9.620 2.54
138 110113815 15.2 10.810 0.890 9.620 5.08
14A 110114A22 16.5 6.060 1.370 8.300 0.953
148 110114822 10.8 6.060 1.370 8.300 4.445
14C 1 101 14C24 6.4 6.060 1.370 8.300 1.905
16A 330216A101 10.2 3.110 5.930 13.320 3.81
17A 330217A108 14 2.580 5.000 9.630 5.398
178 3302178109 7.9 2.580 5.000 9.630 1.270
18A 330218A07 18.4 2.652 7.410 12.600 3.81
188 330218808 19.1 2.652 7.410 12.600 1.27
19C 330219C014 14 2.970 5.000 11.130 1.143
208 110320835 17.1 4.510 0.268 1.210 0.953
21A 110321A32 17.8 1.480 0.268 0.400 0.635
22A 1 10322A43 19.4 2.290 0.362 0.830 0.635
238 1 10323842 20 1.840 0.362 0.660 0.635
24A 1 10324A53 18.7 2.170 0.420 0.910 0.953
248 1 10324854 18.7 2.170 0.420 0.910 0.445
25A 1 10325A53 21.6 1.480 0.420 0.620 0.635

 

 

 

 

 

 

 

 

71

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

Table 4.3: Field Data - All Test Sites (Continued)
Section & Sample Sample Yearly Service Cum Rut
Site Designation Thick ESALS Life ESALS Depth
Number (inches) (millions) (years) (millions) (inches)
268 1 10326888 21.6 2.020 0.670 1.350 1.270
27A 1 10327A82 24.1 1.440 0.670 0.960 0.635
278 1 10327888 24.1 1.440 0.670 0.960 1.397
28A 1 10328A98 20.3 2.070 1.420 2.940 0.635
29A 1 10329A98 22.9 2.160 1.420 3.070 1.270
30A 110330A124 21.6 1.845 1.830 3.185 0.635
318 1103318123 20.3 1.710 1.830 2.970 0.635
31C 110331C124 21 1.710 1.830 2.970 2.54
328 1103328132 10.8 1.845 1.430 3.185 0.953
33A 110333A131 24.1 1.790 0.625 1.120 0.635
34A 110334A145 12.7 4.110 1.420 5.840 0.635
348 1103348153 15.2 4.110 1.420 5.840 2.54
36A 110336A175 15.9 3.870 1.416 5.336 1.270
37A 110337A159 21.3 1.770 0.350 0.620 0.318
38A 110338A176 14.6 3.380 0.830 2.800 1.207
388 1103388175 16.5 3.380 0.830 2.800 2.413
42A 1 10342A249 10.2 4.780 1.250 5.970 1.270
43A 110343A265 7.6 4.360 0.500 2.180 0.953
44A 1 10344A314 21 1.575 10.830 9.795 0.635
45A 110345A312 20.95 3.160 10.830 19.710 0.635
458 1103458312 20.6 3.160 10.830 19.710 1.080
45C 110345C313 21 3.160 10.830 19.710 3.810
468 1 103468364 21 1.420 1 1.670 10.185 1.905
478 1 103478365 20.3 2.600 1 1.920 18.610 1.270
47C 1 10347C364 20.3 2.600 1 1.920 18.610 4.128
48A 1 10348A414 21.6 2.560 10.420 14.920 1.270
488 1103488415 21.6 2.560 10.420 14.920 3.810
49A 110349A4 l 8 21.6 3.420 10.500 19.500 5.715
51A 220351A2905 5.1 0.590 1.670 1.030 0.635
52A 2203 52A272 1 5 12.1 1.270 3.500 3 .420 1.270
528 2203 52827225 9.2 1.270 3.500 3 .420 0.508
53A 220353A2061 10.8 0.510 3.080 1.250 0.635
54A 220454A1213 5.1 0.900 7.420 4.300 0.635
55A 110155A453 1 1.4 0.670 1.250 0.830 0.889
58A 110458A14 5.1 2.880 1.580 4.550 1.016
58C 110458C19 14.9 2.880 1.580 4.550 1.842
59A 110459A46 11.4 2.820 1.580 4.450 3.810
598 110459847 8.9 2.820 1.580 4.450 0.635
61A 330261A014 5.1 2.970 6.000 11.130 4.128

 

 

 

 

 

 

 

 

72

of the tests are mentioned. For details of the test procedure, the reader is referred to the

"Annual Book of ASTM Standards" Volume 4.03.

As noted earlier, all pavement cores were obtained by a four person research team.

The cores were then transported to the MCE laboratories and they were:

1. Inspected to document the possible existence of distress such as cracking and

stripping, and to determine the thickness of the AC course.

2. Subjected to specific gravity tests to determine the bulk specific gravity (ASTM

D-2726) before handling.

3. Washed, dried and oven heated upto the desired temperature to avoid breaking of

coarse aggregate during the process of preparation for extraction test. The material

was then subjected to extraction tests. The purpose of the extraction tests are to

determine:

a) The percent asphalt, fine, sand, and coarse aggregate contents.
b) The penetration of the recovered asphalt cement.

c) The top size and gradation of the aggregates.

d) The coarse aggregate angularity.

A brief description of each test procedure and data reduction is presented

in the next sections.

4.2.1 Bulk Specific gravity Tests (ASTM-D2726-90)

4.2.1.1 Definition

The ASTM D-2726 standard test procedure defines bulk specific gravity as the

ratio of the mass of a given volume of material at 25° C to the mass of an equal volume

of water at the same temperature.

73
4.2.1.2 Summary of Test Method

The specimen is immersed in a water bath at 250 C (770 F). The mass under water
is recorded, and the specimen is taken out of the water, blotted quickly with a camp
towel, and weighed in air. The difference between the two masses is used to measure the
mass of an equal volume of water at 25° C. Correction factors are provided for converting
the mass of water to that of the reference 250 C if from a practical point of view the
weighing was done at different temperatures.

This test method provides guidance for determination of the oven dry or
thoroughly dry mass of the specimen. The bulk specific gravity is calculated from these
masses. Then the density is obtained by multiplying the specific gravity of the specimen
by the density of the water. L
4.2.1.3 Significance and Use

This test method is useful in calculating the percent air voids, as given in ASTM
D 3203, and the unit weight of compacted dense bituminous mixtures. These values in
turn may be used in determining the relative degree of compaction.

Since specific gravity has no units, it must be converted to density in order to do
calculations that require units. This conversion is made by multiplying the specific
gravity at a given temperature by the density of water at the same temperature.
4.2.1.5 Sampling

The samples were obtained in accordance with ASTM D979. The thickness of
specimens be at least one and one half times the maximum size of the aggregate.
Pavement specimens were taken from pavements with a diamond drill. Care was taken to

avoid distortion, bending, or cracking of specimens during and after removal from

74

pavements . Specimens were stored in a safe, cool place. Specimens were free of foreign
materials such as seal coat, tack coat, foundation material, soil, paper, or foil.
4.2.1.6 Procedure
4.2.1.6.] Specimens That Contain Moisture:-

Mass of Specimen in Water - Immerse the specimen in a water bath at 25° C (77°
F) for 3 to 5 min then weigh in water. Designing this mass as C. If the temperature of the
specimen differs from the temperature of the water bath by more than 2°C (36°F), the
specimen shall be immersed in the water bath for 10 to 15 min.

Measure the temperature of the water and if different from 25° : 1°C (77° : 18° F)
a correction to the bulk specific gravity to 25° C must be made in accordance with
procedure. 5

Mass of Saturated Surface Dry - After removal from the water, dry the surface
of the specimen by blotting quickly with a damp towel and then weigh in air. Designate
this mass as 8.

Mass of Oven-Dry Specimen - Oven dry the specimen to constant mass at 110°
C i5° C (230° F) (15 to 24 h is usually sufficient). Allow the specimen to cool and weigh
in air. Designate this mass as A.
4.2.1.6.2 Thoroughly Dry Specimens:

Mass of Dry Specimen in Air - Weigh the specimen after it has been standing in
air at room temperature for at least 1 h. Designate this mass A.

Mass of Specimen in Water - Use the same procedure as described above.

Mass of Saturated Surface Dry - Use the same procedure a described above.

75
4.2.1.7 Calculations

Calculate the bulk specific gravity of the specimen as follows:
Bulk sp gr = A/(B - C) (4.1)
where:
A = mass of the dry specimen in air (gms);
(B - C) = mass of the volume of water for the volume of the specimen at 25°C;

8

mass of the saturated surface-dry specimen in air (gms); and

C

mass of the specimen in water (gms).
After the test, the percent air voids in the AC mix were calculated by using the
following equation. It should be noted that because of lack of data from NHA regarding

theoretical maximum specific gravity (Gmm) of materials, a value of 2.56 was assumed.

Gmm-Gmb
Pal = 100 [ ------------- ] (4.2)
Gmm
Where
Pa = the air voids in the compacted asphalt mixture, percent of total
volume;
G = the maximum theoretical specific gravity of the paving mixture;

mm

Gmb = the bulk specific gravity of the compacted mixture.

P

mm

Gm... = (4.3)

 

Ps/Gse + Pb/Gb

Where

76

PM = percent by weight of total loose mixture =100;

P5 = aggregate content, percent by total weight of mix;
Pb = asphalt content, percent;

Gs, = effective specific gravity of aggregate;

Gb = specific gravity of asphalt.

4.2.2 Extraction Tests

After the determination of the maximum theoretical specific gravity, the samples
were subjected to extraction test in accordance with the ASTM D 2172 standard test
procedure. The tests were conducted to determine the AC mix composition. This test
separates the asphalt cement in the AC mix from the aggregate. Thus, the AC mix
composition such as, the percent asphalt, fine, sand and aggregate contents can be
determined. It should be noted that, the test results may be affected by the age of the
asphalt mix. Older mixes tend to yield slightly lower bitumen content due to aggregate
absorption. It is difficult to remove all the asphalt when some aggregate types are used
and some chlorides may remain within the mineral matter affecting the measured asphalt
content. Nevertheless, Methylene Chloride was used as solvent and bitumen content was
established by difference from the mass of the extracted aggregate, moisture content, and
mineral matter in the extract. The bitumen content (see Table 4.4) was expressed as
percent by the weight of the moisture-free mixtures as follows:

(WI-W2) - (W3-W4)

AC =[ ]x100 (4.4)
(W1-W2)

 

77

AC = asphalt content (percent by total weight of mix);
W1 = mass of test portion;

W2 = mass of water in test portion;

W3 = mass of the extracted mineral aggregate; and
W4 = mass of the mineral matter in the extract.

After extraction of the bitumen from the AC mix, the recovered bitumen, the
aggregate and the fine material were further tested to determine the recovered asphalt

penetration and the gradation and angularity of the aggregate.

4.2.3 Recovered Asphalt Penetration

The ASTM D5-86 standard test procedure was used to measure the penetration
(consistency) of the recovered bituminous material. Higher values of penetration indicate
softer asphalt or lower viscosity. Penetration is defined as the consistency of a bituminous
material expressed as the distance in tenths of a millimeter that a standard needle
vertically penetrates a sample of the material under known conditions of loading, time,

and temperature. The recovered asphalt penetrations are listed in Table 4.4.

4.2.4 Sieve Analysis
The particle size distribution of the aggregate and sand obtained from the
extraction tests was determined by using sieve analysis in accordance with the ASTM C

136-84a standard test procedure. The tests were conducted by passing a weighed sample

78

of dry aggregate through a series of sieves of progressively smaller openings. Table 4.6

lists the percent by weight of aggregate passing through each sieve.

4.2.5 Aggregate Angularity

Since, there is no standard test procedure for determining the coarse aggregate

(retained on sieve number 4) angularity, the number of crushed faces of an individual

aggregate particle was used as a measure of the angularity. Numbers ranging from 1

(rounded and subrounded) to 5 (crushed on all four faces) were assigned to the aggregates

depending on the number of crushed faces. The test procedure was devised by Hamid

Mukhtar during his Phd research (6) and it consists of the following steps:

1.

A 300 gram aggregate sample was obtained irrespective of the amount of the
coarse aggregates in the AC mix.

The aggregates were then divided into different piles depending on the number of
crushed faces. Each pile was inspected and an angularity number was assigned
according to the number of crushed faces of the aggregates. For example, the pile
which consists of aggregate particles crushed on only one face (one side) was
given an angularity of 2. Likewise, the pile that consists of aggregate particles
crushed on all four faces was assigned an angularity number of 5.

The weight of each pile as a percent by weight of the 300 grams sample was
determined.

The angularity of the aggregate sample was then determined as the sum of the
percent by weight of each aggregate pile times the assigned angularity of each pile

and divided by the total weight of 300 grams.

79

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

Table 4.4 : Results of extraction, penetration and Bulk specific gravity tests
for AC course of the cored pavement sections.
Sample Sample Sample Rec Percentage
Designation Thick Sp th Pen AC Sand Fine Agg
Number (cm)
11043A156 8.9 2.445 34.333 4.340 39.747 3.434 52.479
110443146 8.9 2.449 34.333 4.120 42.168 5.149 48.563
11044C159 9.8 2.418 83.333 4.230 43.882 3.409 48.479
11045A13O 1 1.4 2.438 24.667 3.820 49.937 3.549 42.694
11045C13O 10.2 2.234 24.667 4.370 49.039 3.720 42.871
1 1046A121 8.9 2.447 34.667 4.000 40.272 2.736 52.992
110473116 10.2 2.468 25.333 4.810 46.691 2.770 45.729
11047C116 5.1 2.470 95.000 6.170 42.711 2.834 48.285
110483102 5.1 2.485 95.000 7.380 45.563 5.241 41.817
1 1049A98 10.8 2.425 33.000 5.390 45.820 3.945 44.845
1 1049398 16.8 2.454 55.000 5.550 42.975 5.535 45.940
110410A97 12.7 2.481 33.000 5.580 36.484 3.796 54.140
110410397 11.4 2.438 55.000 5.570 44.911 3.891 45.629
110112A37 11.4 2.356 34.000 3.700 46.503 1.839 47.957
1 10112C44 7.6 2.395 65.333 3.400 29.917 2.763 63.920
110113A15 15.2 2.530 82.667 4.890 38.492 2.528 54.090
110113315 15.2 2.535 82.667 4.930 38.409 2.402 54.259
1 10114A22 16.5 2.367 46.000 4.850 36.623 0.866 57.661
110114322 10.8 2.554 96.000 6.860 41.554 2.366 49.215
1 101 14C24 6.4 2.427 96.000 6.470 43.856 2.862 46.812
330216A101 10.2 2.472 98.000 4.600 40.478 2.477 52.444
330217A108 14 2.535 91.333 5.820 31.029 3.246 59.906
3302173109 7.9 2.429 68.000 5.510 37.900 3.275 53.315
330218A07 18.4 2.349 62.667 5.330 43.310 2.353 49.007
330218308 19.1 2.333 42.667 4.600 40.941 3.403 51.056
330219C014 14 2.384 33.000 3.160 40.227 2.140 54.473
110320335 17.1 2.367 72.333 3.570 29.373 3.983 63.075
110321A32 17.8 2.401 72.333 3.840 29.608 3.856 62.696
110322A43 19.4 2.390 44.667 4.340 35.251 2.879 57.530
1 10323342 20 2.400 44.667 4.360 38.026 2.917 54.697
1 10324A53 18.7 2.315 36.000 3.920 44.082 2.056 49.942
110324354 18.7 2.367 36.000 3.480 34.960 1.216 60.344
110325A53 21.6 2.340 36.000 4.150 40.899 3.834 51.117
1 10326388 21.6 2.439 43.000 5.070 42.557 3.873 48.500
110327A82 24.1 2.350 34.667 3.400 43.547 4.038 49.015
1 10327888 24.1 2.382 34.667 4.470 37.992 4.070 53.468
110328A98 20.3 2.410 35.000 4.820 45.705 4.331 45.144

 

 

 

 

 

 

 

 

 

80

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

Table 4.4 : Results of extraction, penetration and Bulk specific gravity tests

for AC course of the cored pavement sections (Continued)
Sample Sample Sample Rec Percentage
Designation Thick Sp th Pen AC Sand Fine Agg
Number (cm)
110329A98 22.9 2.411 35.000 4.040 44.525 4.472 46.963
110330A124 21.6 2.419 42.333 4.510 41.280 4.077 50.132
1103318123 20.3 2.472 42.333 4.860 43.156 3.473 48.512
110331C124 21 2.454 46.333 5.220 45.314 3.422 46.044
1103328132 10.8 2.429 40.333 5.020 43.748 4.445 46.787
110333A131 24.1 2.356 32.000 4.270 44.495 3.475 47.760
110334A145 12.7 2.411 46.000 4.040 41.570 4.165 50.225
1103343153 15.2 2.476 92.333 4.640 32.441 2.212 60.706
110336A175 15.9 2.450 46.000 4.620 28.709 3.930 62.741
110337A159 21.3 2.323 42.000 4.260 43.906 4.155 47.679
110338A176 14.6 2.259 65.000 4.450 38.583 2.303 54.664
1103388175 16.5 2.464 94.000 4.710 36.382 2.497 56.412
110342A249 10.2 2.347 67.000 3.710 33.923 2.754 59.613
110343A265 7.6 2.313 66.333 3.030 32.068 2.987 61.915
110344A3l4 21 2.311 44.000 4.160 39.007 4.524 52.309
110345A312 20.95 2.348 74.000 4.910 40.575 4.098 50.417
1103458312 20.6 2.384 92.333 5.830 40.107 4.106 49.955
110345C313 20.95 2.516 86.667 5.740 40.786 2.969 50.505
110346A361 20.95 2.323 43.000 4.820 43.088 3.940 48.152
1103468364 20.96 2.360 95.000 4.750 38.738 3.648 52.864
1103478365 20.3 2.404 85.000 4.440 36.638 3.478 55.444
110347C364 20.3 2.525 95.000 5.250 34.394 3.373 56.983
110348A414 21.6 2.431 52.667 5.180 30.997 4.219 59.604
1103483415 21.6 2.522 52.667 5.260 35.471 2.975 56.295
110349A418 21.6 2.536 83.667 5.310 33.204 2.746 58.739
220351A2905 5.1 2.510 35.667 5.220 38.433 3.668 52.679
220352A27215 12.1 2.390 34.667 4.220 28.983 3.850 62.947
220352327225 9.2 2.351 34.667 3.330 35.314 4.128 57.229
220353A2061 10.8 2.431 68.333 5.190 42.181 4.722 47.907

 

 

 

 

 

 

 

 

 

81

Table 4.4 : Results of extraction, penetration and Bulk specific gravity tests

for AC course of the cored pavement sections (Continued)

 

 

 

 

 

 

 

 

 

 

 

 

 

 

Sample Sample Sample Rec Percentage

Designation Thick Sp th Pen AC Sand Fine Agg
Number (inches)

220454A1213 5.1 2.438 36.667 4.530 51.458 4.172 39.840
110155A453 11.4 2.424 64.333 4.730 42.776 3.858 48.635
110458A14 5.1 2.265 82.667 2.700 32.946 3.824 60.530
1 10458C19 14.9 2.432 82.667 3.810 29.280 3.251 63.659
110459A46 1 1.4 2.533 66.333 4.950 40.444 2.813 51.793
1 10459847 8.9 2.408 66.333 4.870 38.1 19 2.968 54.043
330261A014 5.1 2.526 64.333 5.650 39.155 4.255 50.940

 

 

 

 

 

82

The above procedure implies that the angularity of the coarse aggregate is
independent of the amount of the coarse aggregate in the AC mix. That is two samples
with 20 and 50 percent coarse aggregate contents may yield the same angularity. The
drawback of this procedure is that, asphalt mixes with low coarse aggregate contents may
have the same aggregate angularity as those mixes made with high coarse aggregate
contents. At low coarse aggregate contents, the aggregate particles float in the sand
matrix which causes separation of the particles. Hence, the aggregate interlocking and
friction are decreased. This scenario implies that care should be taken when analyzing
the effects of the coarse aggregate angularity on the AC mix performance. The aggregate
angularity must be considered in view of the percent aggregate contents in the mix. That
is, higher coarse aggregate content in the mix leads to more mobilization of the aggregate
angularity. Nevertheless, the test procedure and the method of calculating the angularity
were adopted to minimize the dependency of the angularity on the percent aggregate.
This procedure of calculating aggregate angularity follows different approach than those
used by other researchers. In their test method, the aggregate angularity is defined by
three categories (rounded, subrounded, and angular). The sample angularity is calculated
as the weighted average angularity of the three categories based on their weights. The
method does not differentiate between aggregates crushed on 1, 2, 3, or 4 faces.

The angularity of sand size particles (passing standard sieve number 4 and
retained on sieve number 200) can be determined by procedure adopted by the Michigan
transport department as standard test method (MTM) 118-90. The test procedure

however, requires a minimum sample size of 1500 grams which was not available in this

83

study. Hence, the sand angularity was not determined and it is not considered in any
further analysis.

Finally, in all test sites the coarse aggregates were crushed on either 3 or 4 faces.
Because of this consistency (lack of variability in the aggregate angularity ), the data was

not included in the study.

84

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

Table 4.5 : Results of Sieve Analysis for AC Course
Sieve Size (mm)
Sec.| 25 J 19 I 12.5 I 9.5 | #4 l #8 I #50 [#100 [#200 I -200 [Gravel
Percent Retained

1A 0.00 0.00 12.35 12.87 26.09 13.23 24.74 3.78 1.73 5.18 64.54
3A 1.82 3.12 16.62 9.70 23.60 11.92 22.33 4.76 2.54 3.55 66.78
43 0.00 0.00 17.90 11.14 21.61 8.87 24.96 7.17 2.98 5.35 59.52
4c 1.65 4.83 12.79 7.91 23.44 12.18 25.56 5.53 2.55 3.52 62.80
5A 2.12 1.34 9.27 7.7 23.96 15.94 29 4.54 2.44 3.65 60.33
53 6.78 10.02 6.85 21.18 15.08 28.09 5.2 2.91 3.84 59.91
6A 2.01 6.37 14.82 9.82 22.18 11.15 22.91 5.13 2.76 2.81 66.35
68 2.38 4.4 15.64 8.79 17.41 12.22 25.79 4.7 4.1 4.51 60.84
78 2.75 21.04 9.32 14.93 12.63 25.91 8.04 2.47 2.87 60.67
7C 3.77 18.37 10.8 18.52 11.46 25.51 6.38 2.17 2.98 62.92
88 0.46 12.96 10.73 20.4 9.11 30.51 7.41 2.7 5.67 53.66
9A 6.54 11.86 5.53 23.47 12.8 27.81 5.54 2.28 4.13 60.20
98 2.26 2.49 5.4 8.77 29.72 14.23 21.55 6.8 2.92 5.82 62.87
10A 0.79 2.58 19.46 12.37 22.14 8.31 22.99 5.07 2.27 3.98 65.65
108 3.72 10.55 7.72 26.33 12.25 28.03 4.93 2.35 4.05 60.57
11A 3.46 5.6 25.32 10.69 16.58 7.41 10.38 5.9 3.13 3.51 69.06
11C 2.46 9.76 27.67 7.99 14.81 8.31 19.08 3.99 2.38 3.52 71.00
12A 1.31 9.57 19.44 6.19 13.29 9.57 30.82 4.03 3.87 1.88 59.37
12C 6.41 27.76 11.01 20.99 8.3 18.74 2.49 1.44 2.82 74.47
13A 1.1 7.5 20.02 9.72 18.98 8.73 24.88 4.54 1.9 2.6 66.05
138 8.16 20.67 9.38 19.31 10.98 24.16 3.35 1.49 2.47 68.50
14A 4.72 20.32 12.12 23.44 10.23 20.75 6.56 0.95 0.89 70.83
143 3.1 3.29 21.07 9.49 15.89 9.04 29.3 3.385 2.89 2.41 61.88
14C 5.32 19.36 8.75 16.62 9.6 29.91 3.99 3.46 3.06 59.65
16A 1.75 19.92 11.04 22.73 11.73 23 4.5 2.76 2.53 67.17
17A 6.86 4.55 17.34 13.58 21.66 9.13 15.15 5.93 2.39 3.36 73.12
178 1.36 19.2 11.85 24.47 11.25 17.58 6.3 4.56 3.41 68.13
18A 15.99 10.41 25.87 12.71 24.24 6.47 1.85 2.41 64.98
188 0.53 15.53 11.06 26.91 13.32 19.98 7.5 1.67 3.53 67.35
19C 18.59 13.22 24.44 10.96 20.36 7.6 2.62 2.17 67.21
208 5.78 11.76 16.28 8.04 23.55 9.47 13.67 4.57 2.75 4.08 74.88
21A 9.49 4.19 17.82 7.18 26.52 9.9 13.27 5.02 2.6 3.96 75.10

 

 

 

 

 

 

 

 

 

 

 

 

 

85

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

Table 4.5 : Results of Sieve Analysis for AC Course (continued)
Sieve Size (mm)
Sec. I 25 I 19 I125 I 9.5 L#4 I #8 I #50 [#100 I #200 I -200 [Gravel
Percent Retained

22A 4.36 6.25 14.1 9.21 26.22 11.24 17.82 4.87 2.92 2.97 71.38
238 1.22 10.01 13.53 8.74 23.69 10.78 18.82 6.65 3.51 3.02 67.97
24A 1.64 5 19.33 6.29 19.72 16.67 18.48 5.49 5.24 2.09 68.65
248 5.39 13.25 16.44 5.39 22.05 8.25 18.36 5.86 3.75 1.23 70.77
25A 3.33 3.77 13.07 8.42 24.74 13 24.06 3.61 2 3.96 66.33
268 2.19 7.63 16.05 4.41 20.81 14.02 22.11 5.62 3.08 4.05 65.11
27A 3.41 4.12 11.75 3.81 27.65 11.8 23.4 6.08 3.8 4.15 62.54
278 6.8 16.64 5.29 27.24 11.08 21.32 4.78 2.59 4.26 67.05
28A 1.6 14.16 7.44 24.23 11.57 27.68 6.01 2.76 4.51 59.00
29A 0.61 15.38 7.56 25.39 12.16 25.37 5.97 2.9 4.61 61.10
30A 5.67 14.06 8.66 24.11 11.13 22.81 6.1 3.19 4.22 63.63
318 1.03 6.51 11.07 6.6 25.78 9.92 25.01 7.08 3.35 3.64 60.91
31C 1.23 3.56 11.31 5.41 27.07 12.53 25.3 6.9 3.08 3.56 61.11
32A 3.62 7.09 12.87 14.8 14.92 19.06 18.94 3.48 1.88 3.3 72.36
328 5.83 13.5 14.24 15.82 17.25 21.28 5.04 2.82 4.68 66.64
33A 3.27 16.05 5.47 25.1 13.34 24.16 5.62 3.36 3.59 63.23
34A 4.92 20.25 10.37 16.8 15.35 20.65 4.83 2.49 4.31 67.69
348 5.9 12.85 17.12 10.37 17.42 10.13 14.85 5.75 3.29 2.3 73.79
36A 10.84 9.32 12.42 11.77 21.43 8.58 14.3 4.7 2.52 4.11 74.36
37A 5.78 2.45 13.83 5.02 27.72 16.26 22.83 3.98 2.79 4.31 71.06
38A 4.1 5.68 17.88 9.15 20.4 11.05 17.24 8.37 3.72 2.36 68.26
388 2.91 5.82 23.43 9.31 17.73 11.36 16.49 7.18 3.15 2.6 70.56
42A 2.98 22.36 11.92 24.65 12.67 16.18 4.22 2.16 2.82 74.58
43A 1.51 5.27 21.61 12.21 23.25 12.21 16.9 2.44 1.52 3.05 76.06
44A 1.18 3.35 18.43 9.98 21.64 10.85 20.52 5.8 3.53 4.69 65.43
45A 5.26 18.93 9.04 19.79 9.3 22.72 7.3 3.35 4.3 62.32
458 6.81 18.03 7.61 19.72 9.22 21.85 8.03 3.49 4.33 61.39
45C 4.15 20.15 8.88 20.4 9.57 22.55 7.19 3.96 3.12 63.15
46A 4.5 17.23 8.24 20.62 8.98 22.75 10.07 3.47 4.12 59.57
468 5.53 19.15 9.7 21.12 8.71 21.77 6.88 3.31 3.8 64.21
478 2.33 6.33 18.2 9.3 21.86 11.26 17.7 6.01 3.37 3.62 69.28
47C 2.38 8.83 17.63 9.63 21.67 9.22 18.9 5.46 2.72 3.55 69.36
48A 2.45 5.79 26.8 10.3 17.52 7.57 18.77 3.32 3.03 4.42 70.43
488 7.62 8.05 20.04 8.35 15.36 8.2 20.7 4.57 3.97 3.1 67.62
49A 4.96 29.67 9.45 18.39 6.96 21.15 4.29 2.3 2.87 69.43
50A 1.15 0.64 18.22 11.94 26.83 8.17 20.79 4.31 2.79 5.12 66.95
51A 2.81 12.52 12.83 27.42 10.38 23.47 3.92 2.18 3.84 65.96

 

 

 

 

 

 

 

 

 

 

 

 

 

Table 4.5 :

86

Results of Sieve Analysis for AC Course (continued)

 

Sieve Size (mm)

 

Sec. I 25 I 19

L125 I 9.5 I #4 I #8 I #50

I #100 I #200 I -200 [Gravel

 

 

 

 

 

 

 

 

 

 

 

 

 

Percent Retained
52A 5.27 30.16 13.74 16.55 7.23 17.58 3.52 1.93 3.98 72.95
523 4.44 18.42 10.44 25.9 12.12 18.96 3.56 1.89 4.25 71.32
53A 1.06 21.32 10.29 17.86 9.07 25.42 7.47 2.53 4.95 59.60
54A 1.49 11.22 9.7 19.32 11.43 28.21 11.72 2.54 4.35 53.16
55A 1.08 14.45 11.82 23.7 11.08 26.23 5.37 2.22 4.02 62.13
58A 14.29 15.55 32.37 15.63 13.87 2.81 1.55 3.9 77.84
58C 14.35 3.25 19.42 10.58 18.58 8.6 16.33 3.5 2.01 3.35 74.78
59A 19.81 12.65 22.03 15.57 16.84 18.19 1.95 2.93 70.06
598 1.72 1.12 20.02 12.98 20.97 10.6 17.73 8.88 2.86 3.08 67.41
60A 3.58 23.18 10.2 20.47 10.93 22.02 3.35 2 4.22 68.36
61A 4 20.38 10.89 18.72 6.84 22.95 9.2 2.51 4.47 60.83

 

 

 

 

 

 

 

 

 

 

 

 

 

CHAPTER 5

ANALYSIS AND DISCUSSION

5.1 General

The response of a pavement structure to wheel loads can be divided into elastic,
viscoelastic, and plastic. Although the three responses are dependent on the material
properties, the elastic one is time independent while the viscoelastic and plastic responses
are time dependent. The plastic response manifests itself in the form of rutting and fatigue
cracking.

Like most other solids, asphalt concrete pavements exhibit a greater variety of
mechanical behavior than liquid or gases. It is extremely difficult to arrive at a single set of
equations that realistically describes or models the interplay of the elastic, viscoelastic, and
plastic responses under a combined set of stresses and boundary conditions. Even if such
equations could be developed, they would be far too complex for the practical analysis of
stresses and strains induced in the pavement structure. To simplify the problem, in the early
work on plastic solids, the yield condition and plastic flow rule were treated as independent
ingredients of theory of plasticity. For example, an experimental based yield condition was
advanced by Tresca (59). Later, Saint Venant (60) and Levy (61) adopted Tresca's condition
and developed a plastic flow rule whose form was inspired by the theory of elasticity. In his
theory of plasticity, Von (62) retained the plastic flow rule and modified (for the purpose of

mathematical convenience) the Tresca yield condition.

87

88

Recently, the plastic flow rule and yield condition for asphalt surfaced pavements
were further simplified and rut models were developed. The models are based on a single
variable, the compressive strain. The vertical compressive strain is calculated by using
elastic layer theory. Examples of these models are those developed by the AASHTO/ARE

(63) and Majidzadeh and Ilve (64). The main limitations of the models are that:

1. They are independent of the plastic properties of the paving materials.

2. They cannot be used to assess the effects of material properties on the rut potential
of asphalt pavements.

3. Their accuracy is poor at best.

Numerous statistical rut depth models for AC surfaced pavements have also been
developed. The models express the rut depth in terms of the asphalt-aggregate mix
properties and compositions. These models are presented in Chapter 2 of this thesis. It
should be noted that the rut depth of asphalt pavements are fimctions of not only the AC
layer properties but also the properties of the base and subbase layers, the characteristics of
the roadbed soils, the traffic load and volume, relief of the area and the environmental
factors. In Pakistan, most of the asphalt pavement network is experiencing premature
rutting and it is mainly contributed by AC layer.

The objectives of this study (as explained in chapter 1) are to determine the asphalt
mix variables that affect the structural performance (rutting) of flexible pavements and to
recommend changes in the existing asphalt mix design procedure and standard
specifications in Pakistan. To accomplish these objectives, the outputs of this study must

include relationships between pavement performance and asphalt mix variables such that

89

the sensitivity of the pavement performance to the various asphalt mix variables can be

assessed. Therefore, the data analysis procedure to be employed, must account for the

variability of the asphalt mix and must be capable of:

1. Providing a good description of the sensitivity of the response of the dependent
variable (rut) to the input parameters (independent variables).

2. Developing realistic models whereby the relationships between the dependent and
independent variables are reasonable and have good engineering interpretations.

3. Predicting pavement responses in terms of rutting potentials for the range of the
given variables.

Two types of analysis can be employed; mechanistic and statistical. The former is
typically based on existing theory (e.g., elastic, plastic, viscoelastic) and it requires
substantial inputs regarding material properties. Because of the limited resources, such
inputs cannot be obtained. Further, the resulting mechanistic models are not practical and
they cannot be used by Highway Agencies. Consequently, a statistical type analysis was
selected in this study. The inputs to the statistical models include some of the engineering
and physical properties of the AC materials as well as traffic loading, volume, grade, and
pavement service life (age).

The main advantage of the statistical analysis is its simplicity and low cost. The
main disadvantage is that the resulting statistical models are applicable only within the
range of data from which the models have been developed. This disadvantage can be
minimized if the range of material properties used in developing the statistical model

represents the spectrum of the properties used in the field. Therefore, the selection of the

9O

experiment is very crucial to the success of the analysis. The experiment should include

pavement sections that:

1.

Have a representative range of the properties and variability of the asphalt mixes
used in the construction of various pavement sections;

Experience light, medium, and heavy traffic;

Are located in the various environmental regions;

consist of various cross-sections.

The details of the experiments in this study are presented in Chapter 4. The

statistical package (computer program) used in the study is "Statistical Package for Social

Sciences (SPSS)". The reasons for this selection include (4):

1.

2.

The simplicity and user fiiendliness of the program.

The various options that are embedded in the program which allow the users to
examine the various statistical parameters of the output models.

Its graphic capability whereby each input and output variable can be plotted and
examined for engineering interpretation.

The capability of the program to access the data through a common spreadsheet
and/or data base.

The flexibility of the program which allows the users to input the desired form of
equations relating the dependent and independent variables.

Most of the statistical procedures are generally based on the least squares criteria

(minimization of the sum of the square of the differences between the predicted and the

observed data). These procedures are simple and easy to use provided that the proper

9l

relationship between each independent variable and the dependent one is specified. In this

study, emphasis were laid so that the relationship would:

1. Have the proper engineering interpretations.
2. Be representative of the observed trend between the dependent and independent
variables.

Several aspects relative to these issues and some specific comments regarding the

statistical procedure used in this study are presented in the next section.

5.2 STATISTICAL ANALYSIS TECHNIQUES AND ISSUES
The most important issues regarding statistical analysis of engineering data include
(65, 67):
1. The number of variables to be included in the analysis, their significance level, and
their collinearity if any.
2. The form of the equation(s) relating each independent variable and/or clusters of
independent variables to the dependent one.
3. The engineering interpretations of the final correlation equation.
If the only purpose of the statistical analysis is to express the behavior of the response
(dependent) variable in terms of the independent variables and to minimize the residual sum
of squares, then the best suited model is the one that includes all available independent
variables regardless of whether or not, the resulting relationship is realistic. A model having
fewer variables, on the other hand, has the appeal of simplicity, and economic advantage in

terms of obtaining the necessary information. The elimination of some variables from the

92

model, however, is obtained at the expense of biases and loss of predictability of the model.
Thus, the exclusion of any variable with a significant correlation coefficient causes a bias
penalty (65).

Regardless of the number of variables to be included in the final model, the results
of any statistical analysis that is based solely on the least square technique reflects only the
correlational structure of the data being analyzed. This structure may or may not be
representative of the true engineering relationships between the dependent and independent
variables. Another problematic aspect of variable selection is that the relative importance of
a variable as manifested in the sample may not necessarily reflect its relative importance in
the population. Important variables in a sample may appear unimportant in the population
and vice versa. That is, the best set of variables in a sample might not be the best set in the
population. The choice of variables in the final model depends on prior knowledge and
experience and on the variable clusters used to model perfect behavior. Further, statistical
correlation between independent variables (e,g., material properties), and the dependent
variable (e,g., pavement performance in terms of rut resistance) may lead to several possible
outcomes including:

1. Certain material properties (e.g., air void) appear to have specific effects on
pavement performance which can be related to certain observed patterns.

2. Certain material properties appear to have no effect on pavement performance. That
is, regardless of the range of the property and its variation, the pavement

performance is more or less constant for the entire range of that property.

93

3. Variation in the values of the pavement performance appear to have similar pattern
that can be related to variations in the material properties.
Nevertheless, in this study, the statistical analysis were conducted by using the
SPSS computer program. The statistical models (regression equations) were developed by
using four steps. A brief discussion of each of these steps is provided here and a detailed

discussion is contained in the model development (67).

Step 1 Determination of a Simple Correlation Matrix

In this step, a simple correlation coefficient between the dependent and each of the
independent variables was computed and tabulated in a matrix format. Each coefficient was
then examined to determine the trend between the dependent and each of the independent
variables and the significance of the latter on the former. The trends and the significant
levels were also compared to field observations and experience. The simple correlation
matrix also provides information regarding the degree of collinearity between any two
independent variables (a situation when two or more independent variables are correlated
among themselves). A higher degree of collinearity between any two independent variables,
causes more difficulty in separating the effect of each independent variable on the response
variable(66). Because of its importance, the degrees of collinearity between the independent

variables were also examined in steps 2 and 3.

94

Step 2 Computation of Eigenvalues and Condition Indices

The eigen values of a symmetric (square) matrix "A(n*n)" are a set of "n" non-
negative scalars "2.3" such that their product with n non-zero vectors "2,, i =1...n" of the
matrix is the same as the product of the matrix with the z,- vectors. That is "A2,- = 21,2 z". The
eigenvalues are typically used to examine the collinearity between the independent
variables. In addition, the scaled uncentered cross-products matrix and the decomposition
of the regression variance corresponding to these eigen values were computed. There is an
indication of near dependency of variables, when there is a high proportion of the variance
of two or more coefficients that are associated with the same eigen value (67). Hence, the

condition index which is defined below can be calculated.

Cond. Index =[ eigenvaluemax / eigenvaluei ]°‘5 (5.1)

The presence of collinearity results in small eigen values and consequently, larger
condition index values. Furthermore, the number of large condition index values

corresponds to the number of cases of collinearity between the independent variables.

Step 3 Computation of Variable Tolerance and Variance Inflation Factor (VIF)
Variable tolerance and variance inflation factors (V IF) are other measures of

collinearity between the independent variables. The variable tolerance and the VIP are

defined as (64, 67):

95

Tolerance = (l-Rf) (5.2)

VIF = [ 1/(1-R,-2)] (5.3)
where

R- = the multiple correlation coefficient of the ith independent variable when it is

predicted from other independent variables.
If the tolerance of an independent variable is small, or its VIF is high, then the

variable is collinear with one or more independent variables.

Step 4 Variable Selection Methods

In this step, the independent variables to be included in the statistical model are
selected. Several variable selection methods are available to the user of the SPSS program.
In this study, the stepwise regression method is used. This method identifies a good (not
necessarily the best) set of variables to be included in the development of the statistical
model. The statistical model is constructed by considering one variable at a time that
having the greatest or the least impact on the residual sum of squares of the model. Three
stepwise techniques can be identified (62, 67); forward selection, backward elimination and
stepwise selection. These three techniques are addressed below.
Forward Selection - In the forward selection technique, the first independent variable to be

considered in the development of the statistical model is that which accounts for the largest

96

amount of variation in the dependent variable. At each successive step, the independent
variable (from the independent variable pool) that causes the largest decrease in the residual
sum of squares of the existing model is added. In the absence of any termination rule the
process continues until all the variables are included in the model.
Backward Elimination - In the backward elimination technique, first the statistical model
is developed by including all the independent variables in the pool. At each consequent
step, a variable whose deletion causes the least increase in the residual sum of squares is
dropped. If no termination rule is specified the deletion process continues until only one
independent variable is left in the model.
Stepwise Selection - The stepwise selection technique is basically a forward selection
process that at every step reexarnines the significance of the previously added variables. If
the partial sum of squares of any previously added variable does not meet a minimum
specified criterion, the variable is dropped from the model and the selection process
changes to backward elimination. The variable elimination process continues until all of
the variables remaining in the model meet a minimum specified criterion after which the
forward stepwise selection process is resumed. It should be noted that, at every step, the
significance of all variables relative to the model is reexamined. Hence, any variable that
has been deleted from the model in earlier steps may be added back to the model when it
meets the minimum specified criterion (67).

Neither the forward selection nor the backward elimination techniques consider the
impacts of adding or deleting a variable on the remaining variables in the model. A variable

added to the model in the forward selection technique may become insignificant after other

97

variables have been added. Similarly, the significance of a deleted variable may increase as
other variables are deleted from or added to the model.

The stepwise selection technique was used in this study because the process allows
a better chance of selecting the best independent variables relative to the other two
techniques. Nevertheless, the forward selection and the backward elimination techniques
were occasionally used to compare the resulting statistical models. For each statistical
model, the null hypothesis of no relationship between the dependent and independent
variables were tested. Finally, the residual analysis and the comparison of the predicted

versus observed values were performed.

5.3 ANALYSIS AND DISCUSSION OF RUT DATA

Various Rut Models for asphalt concrete pavement have been established by
numerous investigator based on field performance and laboratory data. The laboratory
based rut model generally predict failure much sooner than is observed in the field.

Rut or plastic deformation of a pavement structure is a function of the construction
practices, the properties and thickness of the AC, base, and subbase layers, the properties of
the roadbed soil, environmental factors and traffic load and volume (1). Unfortunately, the
properties of the base, subbase, and roadbed materials are not available to this study.
Hence, the following assumptions were made at the onset of this analysis.

1. Rutting in flexible can be analyzed as a function of the AC layer and its material

properties.

98

The material properties, layer thickness, construction specifications/practices and

the construction quality remain the same within one site.

The two assumptions are reasonable because of the following reasons:

1.

The pavement sections included in this study and most of the flexible pavement
network in Pakistan are experiencing premature rutting. The rut is mainly
contributed by the AC layer (the rut channel is narrow and deep). Hence, the
properties of the base, subbase layers and the roadbed soil have insignificant effects
on pavement rutting.

In this study, the test sites within one pavement test section number, were selected
from the same job number (a number, used by Highway agencies to identify a
construction/rehabilitation project). It is the practice of Highway agencies, to divide
longer pavement control sections into several smaller job sections for construction
and/or rehabilitation purposes. The pavement thickness design, the material
properties, and the specifications are then kept the same within the same job section
(number) depending on the existing pavement. Thus, any pavement section within
the same job number will “ theoretically “ have similar layer thicknesses and
material properties.

Due to the first assumption, one should expect some bias in the developed rut

model. That is, given the lack of information regarding the properties of the base, subbase

and roadbed soils, their effects cannot be incorporated. Further, the assumption of constant

material properties may not hold. Even though, in the presence of the same specifications

and design, variations due to construction practices are always present in the form of the AC

99

mix manufacturing, placement, and compaction. Nevertheless, the thickness of the AC
layer in the pavement was obtained from cores that were extracted at one or more locations,
and the AC mix properties for each core were determined as presented in the next section.
Thus, the bias in the data is minimized and whatever exists, is mainly because of variation

in the thickness and properties due to construction.

5.3.1 Test Results

A total of 75 six/four inch diameter full depth cores were obtained from cored
sections/site. Each core was carefully examined and defects such as stripping, cracking and
smoothness of the core sides were noted and the thickness of the AC layer in each core was
measured and recorded. Prior to extractions, the bulk specific gravity of each core was
ascertained using ATSM D 2726 standard test procedures. The test result and field data are
listed in Table 5.1.
5.3.2 Determination of Material Properties for a Core Location

Based on the second assumption and due to economic and time constrains, and test
requirements (the materials from more than one core must be combined to obtain the
required amount of material), it was decided to perform one extraction test for each
pavement section. Once again, prior to testing, two or more cores obtained from one
pavement site were combined to yield enough material for a single extraction test. Table
5.1 summarizes the test results and field measurements. These include the sample
designation number, thickness, longitudinal grade, specific gravity, air voids, the

penetration of the recovered asphalt cement, the percent by weight of the AC, fine, sand,

100

and coarse aggregate in the asphalt mix, the service life of the pavement section (the period
in years between construction and coring), the cumulative l8 kip ESALS experienced by
pavement section during the service life and the average rut depth of the core location.
Table 5.2 summarizes the minimum, maximum, average, and standard deviation of each
data entry of Table 5.1.
5.3.3 Statistical Analysis
The objectives of this analysis are to :
1. Determine the influential factors affecting pavement rutting.
2. Recommend changes to the existing asphalt mix design and pavement construction
practices to decrease the rut potential.

Statistical analysis was conducted to relate the measured pavement rut depth to
the AC mix variables, the longitudinal grade and the traffic volume. During the analysis,
several collinearity problems were encountered. For example, by normal practice, higher
angular aggregate contents, better construction practices, thick AC layer, and strict
quality control are commonly used for most pavements designed for higher traffic
volumes. By the nature of this practice, collinearity between several independent variables
(e.g., AC thickness, and traffic volume) exists as shown by the simple correlation matrix in
Table 5.3. For example, the term ESAL (the cumulative l8-kip equivalent single axle load
applications) has positive collinearities with the AC thickness (.1), the percent asphalt

content (.48 ), bulk specific gravity (.35 ) and the penetration (.45 ).

2.. sneaz
2.. 325 342mm 6.5 mu< 6:; :63 88> be ..m 8:: 8:28.83 25
5:0 his“; «:35.— oom .._< 0::— U< 9.5—Sam ~85on

 

sass so. .2 3% £5 \ 22... u E 2.3.

.x. $5852
5%: 2.. 342mm 3.23 a: 8.— 82.5 be ..m 3.2.; 8:838: 25

 

Aegeanoovuo—mfian 3.8 .5.“ Sue and \ Eur.— " mm 932,—.

103

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

23 _ 8.8 8.8 8.: :8 :8 8.8 :8 8 8.: 82.: 8.: 82.848: <2.
:3. o 8.: 8.: 8.: 8.8 8.8 8.8 88 8 8.8. 84.: 8.: 2888: 8:.
88: 8 8.: 8.: 8.: 8.8 8:. 8.: :8 8 88 :8: 8.: 3888: <8.
8:. o 8.: 8: 8.: 8.8 8.8 8.8 8.8 8 8.8 :4: 2: 88:8: on.
88: o :3: 8: 8.: 3.8 8.8 8.8 3:. 8 :8 8...: 2: 885.8: ms.
88: o 2.: 8: 8.: 8.8 8.... :8 8:. 8 8.: 82 8.: 882.8: ms.
:8 o 8:: 8: 8.: 8.8 8.... 8.8 8:. 8 88 82 2: 822.8: <2.
:3 o :8: £8 8.: 8.8 8: 8.9. :8 8 8.8 84.: 8.: 2888: 68
88: o :8: 2.8 8.: 8.8 ::. :2. 88 8 8.8 2.8: 8.8 885.8: 8:.
88 o :8: :8 8.: 8.8 2:. 8.9. 8:. E 82 82 EN :288: <8
8.8 o 8.8 8: 8:: :8 8:. 8.8 2:. a. 8.8 :3 8.: 5.5.3.8: <3.
88 o :8 8:. 88 8.8 8.8 8.8 8.8 8 8.8 :2 8.: 8.288: <8
88: o 8.8 8:. 8: 8.8 :8 8.8 :8 8 :8 2.2 .8: 2.888: <8
:2 8: 8: 8.8 8.8 3.8 on... 8.8 ::. 8 8.8 82 8: 8588: m8
:8: N: 8: 8.8 88 8.8 8.: 8.8 8:. 8 8.: 88.: 3.: 828.8: <8
:8 o 88 8: 8.8 8.2. 2:. 8.8 8:. 8 8.8 82 n: 8288: <8
88: o ...? 8.8. 8: :8 8.8 :8 8.4 s. 8:. 88.: 8.2 8288: <8
2.2 : 8.8 ::. 8: :8 :.~ 3.8 8:. 8 8.8 85.: S: 8588: 82
8.8 o a: ::. 8: 8.8 2:. 8:. 8:. s. :2 :2 Z: 8288: <8
88 o a: 8: 88 8.2. :8 as. 8:. 8 8.: 82 :8 ::<88: <88
88.8 o :8 8: 8: 8.2. 8.4 8.8 8.8 s. :8 88: 8.: 8588: m8
is 828.5 828.5 888 2. is 8:52

£8: 2. 82mm 828: a: 8< 65.: .88 u< 8.: 88> be am 3888 88:88: 88
.3— 330 :50 buao> oomtom .383 .3 2.33m 31 :3. :15 U< can—am notoom

 

Goa—5588383 33 ..on 3.6 and \ 20E " fin «Ssh.

 

me.~ —n.v wvdm nvd mvdn 9.9 hm chm Quin an:

Ax. Lona—=7:
5%: .x. 82mm 88.8: a: 8E 5.: 88> bu 8m :25 5:288:

 

89:3...88238 .8. ..e as. :3 \ :2..— u E as;

105

 

 

 

 

 

 

 

 

 

 

 

 

Table 5.2 : Descriptive Statistics of table 5.1
Variables Mean Std. Deviation
AC°/o 4.66 0.86
AV 5.70 2.79
AC thick 5.73 2.33
Agg 52.40 5.87
value of fine 3.48 0.89
Sand 39.46 5.37
Rec Pen 56.50 23.33
Grade 0.53 0.90
Cumulative 6.89 5.64
ESALS
Rut depth 1 .60 1.29

 

 

 

 

106

In the correlation matrix of Table 5.3, the positive correlation of ESAL to the AC layer
thickness shows that highway agencies in Pakistan, are designing thicker pavements for
higher ESALS. The correlation is not significant which depicts that this principle is not
being followed in totality and that costs are playing a major role.

The positive correlation with the percent asphalt content, recovered asphalt penetration
and negative correlation with the percent aggregate, substantiate the fact that in Pakistan,
better materials are not being used for higher volume roads. Again, pavement decisions are
being driven by cost, not by experiences. It has already been proven (16,19,21,22) that the
performance of AC mixes depends on providing adequate aggregate interlock to distribute
the wheel load. Increase in the percent aggregate, reduces the amount of sand and fine in
the AC mix. This leads to a better aggregate face to face contact and reduces the ball
bearing effect due to the presence of excessive sand and fine in the AC mix.

The above scenario implies that the statistical analysis between the dependent and
independent variables must be conducted on the basis of engineering knowledge rather than
on the basis of statistics alone. This point can be illustrated by examination of the values of
the correlation coefficients between the dependent variables (rut depth) and the independent
variables (material properties, traffic volume, and service life).

Such an examination leads to the following observations:

1. The rut depth increases with increasing the number of cumulative ESAL. This
observation was expected and it is consistent with that reported in the literature. The reason
for this is that pavement rutting is mainly caused by wheel loads and is accelerated by

material and environmental factors. The correlation is significant.

 

107

Table 5.3 : Correlation matrix for the 75 extracted core locations.

 

P.” can

Corr

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

cum no air no an ago and £111.. grad roe rut
coal voids thick qty 0 pen dop

cum 1.00

coals

ac % 0.48 1.00

air -0.35 -0.57 1.00

voids

ac thick 0.10 -0.08 0.12 1.00

can qty 0.35 0.57 -1 00 -0.12 1.00

agq -0.03 —0.39 0.04 0.09 -0.04 1.00

and -0.04 0.24 0.03 -0.08 -0.03 -0.98 1.00

tines -0.05 0.12 0.07 —0.01 -0.07 -0.33 0.17 1.00

grad. -0.13 0.20 -0.31 0.11 0.31 -0.03 0.04 -0.24 1.00

rocpon 0.45 0.39 —0.34 -0.07 0.34 0.27 -O.33 -0.20 0.27 1.00

rut 0.54 0.51 -0.64 -0.03 0.64 0.11 -0.16 -0.27 0.37 0.60 1.00

depth

 

 

 

108

The rut depth increases with increase in the longitudinal grade of the pavement. This
is consistent with the observations (pavements on grade are experiencing higher
ruts). The experience interpretation is that, higher grade causes lower truck speed.
This implies that the pavement is being subjected to load for longer period of time.
Since, the plastic deformation increases with time (creep), lower truck speed causes
higher rut.

The rut depth increases with the increase in the recovered asphalt penetration. This
observation is reasonable and it is consistent with the literature. Softer asphalts
produce softer asphalt mixes with lower resistance to rutting. The correlation is
partially significant.

The rut depth increases with an increase in the asphalt content in the mix. This was
expected and it is consistent with the literature because higher AC contents cause
more lubrication between the aggregate and thus results lower friction. The percent
asphalt content has higher negative collinearity (0.57) with air voids which reflects
that increase in AC % decreases air voids, thus resulting in rutting.

The rut depth increases with increasing percent of coarse aggregate and with
decreasing percent sand contents in the mix. Both observations are not reasonable
and they cannot be supported by engineering principles. The reason for the two
observation is statistical in nature. Higher coarse aggregate contents in asphalt
mixes has lower surface area and hence lower AC contents. It can be seen from

table 5.3 that sand and aggregate contents have collinearities with AC content in the

109

order of 0.24 and 0.39 respectively. The implication of the discussion above is that

the effect of the AC content on rut depth needs to be examined very carefiilly. The

effect in the final statistical model will be affected by this collinearity.

6. The rut depth increases with decreasing percent fine (passing sieve number 200)
content. This observation is not reasonable and it is not consistent with that reported
in the literature. The herein the same as that reported in item 5 for the sand and
coarse aggregate contents.

7. Finally, the rut depth increases with increasing bulk specific gravity. This
observation is partially reasonable. This quantity is directly used to calculate the air
voids. It is also evident from figure . 5.3 that air voids have highest negative
correlation with specific gravity (more specific gravity less of air voids). The air
voids have parabolic/polynomial relationship with rut depth. Upto some level of air
voids rut depth has negative and then positive correlation. Moreover AC % also has
significant negative correlation with air voids which means that higher AC %,
lower air voids.

The above observations indicate that collinearities exist among some independent
variables. Stratification of the data to resolve the collinearity problem is not possible
because of the nature of the field experiment. Hence, some other solution must be found
before balanced statistical and engineering analysis can be conducted.

In order to alleviate the problem and to analyze the rut depth as a function of the
material variables without a significant influence by the independent variables on each

other, the following corrective measures were adopted.

110

1. The collinearity between the sand and aggregate contents. This can be eliminated
by expressing the percent coarse aggregate contents in terms of the fine and sand
contents as follows:

AGG = 100 - (Fine + Sand)
where AGG, Fine, and Sand are the percent aggregate, fine, and sand
contents, respectively.

2. The collinearity between the air voids (AV) and the percent AC content (AC). The
reason for this degree of collinearity was explained earlier. Unfortunately, because
of the nature of this dependency, this collinearity problem cannot be resolved by
expressing one variable in terms of the other. Thus, only one of the variables will be
included in the model.

Some of the collinearity problems cannot be resolved prior to the statistical analysis.

Consequently, several statistical analysis were conducted whereby various transformations

were used to express each independent variable. It was found that the elimination of certain

variable fiom the pool of variables did not cause any significant bias in the resulting model.

Therefore, those variables were not included in the final analysis.

Based on the results of the statistical analyses presented in Tables 5.3 and the above
discussion, and after the elimination of some variables from the pool , statistical analysis
was conducted to obtain a regression equation relating the dependent and independent

variables. This analysis is presented in the next section.

111

5.3.4 Regression Equation

Prior to the commencement of the statistical analysis, the dependent variable (Rut
Depth) was plotted against each of the independent variables. In view of the plots, the trend
between the dependent and each of the independent variables was observed. The observed
trend of each independent variable was then modeled by using several transformations (e.g.
linear, logarithmic, cubical, exponential, parabolic and polynomial). The form that yielded
the lowest PIN value, the highest coefficient of correlation, highest significance level and
lowest standard error in the presence of the other independent variables was selected for
inclusion in the model.

Table 5.4 provides a list of the values of the correlation coefficients for all variables
that satisfied the variable inclusion criteria (PIN = 0.05) in the final model. The variables
included in the model in order of their significance are:

1. the percent air voids (AV) of the AC mix which is included in a polynomial form as

AV and AV;

2. penetration of recovered asphalt in the form of (REC PEN)“;

3. gradient of the road section in the form of (GRADE)2'25;

4. cumulative ESALS (cum esal) that has been experienced by the pavement since
construction or rehabilitation in the form of (ESAL)”.

The degree of collinearity between the independent variables listed in Table 5.4
were examined. It is evident from the results, that the terms “AV2 & AV3 have the highiest
collinearity. Because of this high collinearity, the term AV3 was eliminated from the final

model. Table 5.5 provides a summary of the values of the coefficients of correlation and

112

standard errors for each step of stepwise method. Table 5.6 provides, for each independent
variable, a list of the six eigenvalues and condition indices and the percent of variance
attributed to each eigenvalue. The results indicate that the different terms of air void, (AV
and AV) have high variance proportion values for the 6th eigen value thus indicating
significant multicollinearity. A similar conclusion can also be made based on the tolerance
and VIF values listed in Table 5.7. These diagnostics indicate that a significant degree of
collinearity exist only among the various polynomial terms of the same independent
variable (the percent air void AV). Note that the presence of such collinearity does not
affect the estimated parameter of the other independent variables.

The results of the regression analysis are listed in Table 5.7. All the variables that
are included in the resulting equation satisfy the specified criterion for the variable
inclusion. It can be seen from the Table that all the variables have a "T" significance value
of less than 0.05. Hence, they are statistically significant variables. The results of the
regression analysis listed in Table 5.8 indicates that the standard error (S.E.) of the
resulting model is 0.203, the value of the coefficient of determination (R2) is 0.852 and the
adjusted R2 is 0.840.

The numbers listed under the column designated "B" in Table 5.8 are the regression
constants for the indicated variables. Table 5.8 also provides a list of the regression
constants and the values of the R2 and the standard error obtained from each step of the
stepwise method of the SPSS computer program. This percentage increases, as it should be
expected, with the inclusion of more variables in the model. The final model can be

expressed as follows:

 

113

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

Table 5.4 : Correlation matrix for the final variables.
rut av av’ avI esal‘fl gradefl! rec pen'5
depth
Pearson rut depth 1-00
Correlation av -0-68 1 .00
av2 -0.51 0.96 1.00
avJ -0.51 0.96 1.00 1.00
W— 053 -o.37 -0.3o -o.3o 1.00
gradeus 0.47 -0.32 -0.26 -0.26 -0.05 1.00
rec pens 0.58 -0.32 40.25 -0.25 0.43 0.25 1.00
Table 5.5: Model Summary
Model R R Adjusted Std. Error Durbin-Watson
Square R Estimate
Square (cm)
1 0.680 0.462 0.454 0.955
2 0.846 0.716 0.708 0.698
3 0.898 0.806 0.797 0.582
4 0.908 0.824 0.813 0.558
5 0.923 0.852 0.840 0.516 2.056

 

 

 

 

 

 

 

 

114

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

Table 5.6 : Collinearity diagnostics (eigenvalue, condition indices and the
proportion of the variance attributable to the eigenvalue).
Eigen Cond Variance Proportions
value Index
Model Dimension Constant av av2 rec gradez-25 esal-75-
peas

1 l 1.90 1 0.050 0.050
2 0.10 4.345 0.950 0.950

2 1 2.74 1.000 0.008 0.002 0.004
2 0.25 3.310 0.167 0.001 0.057
3 0.01 16.901 0.824 0.997 0.939

3 1 3 .57 1.000 0.001 0.001 0.002 0.002
2 0.40 3.007 0.008 0.003 0.039 0.026
3 0.02 12.436 0.203 0.123 0.203 0.619
4 0.01 22.212 0.788 0.873 0.755 0.352

4 l 3.73 1.000 0.001 0.001 0.002 0.002 0.010
2 0.94 1.998 0.000 0.001 0.004 0.000 0.593
3 0.30 3.528 0.011 0.002 0.040 0.031 0.329
4 0.02 12.930 0.210 0.111 0.180 0.672 0.033
5 0.01 23.097 0.779 0.886 0.774 0.294 0.035

5 1 4.40 1.000 0.001 0.001 0.001 0.001 0.007 0.007
2 0.94 2.159 0.000 0.001 0.005 0.000 0.494 0.002
3 0.52 2.915 0.000 0.002 0.015 0.002 0.226 0.164
4 0.1 1 6.273 0.036 0.000 0.049 0.049 0.146 0.661
5 0.02 14.825 0.200 0.089 0.127 0.801 0.066 0.114
6 0.006 25.657 0.762 0.908 0.802 0.147 0.062 0.050

 

 

 

 

115

RUT DEPTH = 2.350 - 0.830 AV + 0.0506(AV)2 + 0.193(REC PEN)'5

+ 0.0936(GRADE)“5 + 0.102(ESAL)” (5.4)

R2: .852, SE=.203, F=76. 864, F(,,g,=.000

Based on the statistics of the model (equation 5.4), the following observations were
noted:
1. The null hypothesis (no linear relationship between the independent variables
and rut depth) was rejected and it was concluded that at a probability Fm) of zero,
85.2 percent of the variation of the dependent variable (rut depth) is explained by
the independent variables included in the equation.

2. The largest value of "t" significance is .0007. This implies that all the variables in
the model are statistically significant. In view of the 95% confidence interval
(default value), t significance of the variables be lower than 0.05.

In addition, the following statistics were calculated to study the values of the

residuals and the predicted rut depth values and are listed in Table 5.9.

1. PRED = the unstandardized predicted values;
2. RESID = the unstandardized residual values;
3. ZPRED = the standardized predicted values;

4. ZRESID = the standardized residual values; and
5. D.W.S. = the Durbin Watson statistics.

 

116

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

Table 5.7 : Results of regression analysis and Collinearity diagnostics
TOL and VIF.
Unstand Stand T T sig Collinearity
Coefficients Coefficients Statistics
Models B Std. Beta Tolerance VIF
Error

1 Constant 3.396 0.256 13.28 5.1E-18
av -0.315 0.040 -0.680 -7.80 1513-] 1 1.000 1.00

2 Constant 5.295 0.304 17.43 5.5E-18
av -l.111 0.105 -2.395 -10.62 5.5E-18 0.080 12.56
av2 0.065 0.008 1.788 7.93 8.0E-12 0.080 12.56

3 Constant 2.827 0.506 5.59 4213-07
av -0.972 0.091 -2.096 -10.73 5.9E-18 0.074 13.56
av2 0.058 0.007 1.582 8.26 1.2E- 12 0.077 13.03
rec pen's 0.269 0.048 0.321 5.64 3515-07 0.867 1.15

4 Constant 2.657 0.490 5.42 8.3E-07
av -0.915 0.090 -1.973 -10.22 6.3E-18 0.069 14.39
av: 0.055 0.007 1.497 8.03 5.2E- 12 0.074 13.43
rec pen! 0.253 0.046 0.302 5.48 6.6E—07 0.853 1.17
grade“? 0.067 0.026 0.146 2.63 1.08-02 0.847 1.18

5 Constant 2.35 0.461 5.10 3.1E-06
av -0.83 0.086 -1.789 -9.62 1.4E- l 6 0.064 15.62
av2 0.050 0.006 1.378 7.84 1.6E-1 1 0.072 13.94
rec pen'5 0.193 0.046 0.230 4.19 8.4E-05 0.734 1.36
gradgfl-5 0.093 0.025 0.203 3 .78 3.3E-04 0.770 1.30
688133 0.102 0.029 0.203 3 .54 7.4E-04 0.675 1.48

 

 

 

 

 

 

 

 

 

 

117

The Durbin-Watson statistics (2.056) of equation 5.4 listed in Table 5.5 signifies a
satisfactory non autocorrelation between the residuals. If variables in the model have
autocorrelation, the values as well as the Standard errors of the parameters estimates are
affected. Values of the parameters will be greater than normal and SSE will be under
estimated. Prediction power of the model will be poor. Acceptable level of Durbin Wastin
is from 1.50 to 2.50 (67).

The analysis of residual statistics shown in table 5.9 indicates that this model is
capable of predicting a rut depth fi'om 0.153 inch to 5.097 centimeters. The maximum
standard error of the predicted values of maximum rut depth is 0.295 which is reasonable.
The process of standardization includes averaging of out lying point inclusion of influential
data. In table 5.9, the minimum and maximum values indicate the spread of the values and
standard deviation helps in understanding the nature of the dispersion of the predicted
values.
5.3.4.1 Shortcomings of the Model

Although the rut model, presented in equation 5.4, has a relatively high R2 value
(0.852), but its engineering interpretation is problematic. For example, the model indicates
that pavement rutting is possible even without any traffic load (ESALS = 0). The reason for
this could be inaccuracy of the ESAL data which was obtained from NTRC and NHA of
Pakistan. It should be noted that other forms of the rut prediction equations were also tried.
For example, several forms where, the ESALS was a multiplier or in logarithmic functions,
were used. Unfortunately, the resulting statistical equations were very poor with R2 values

of less than 0.3.

118

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

Table 5.8: Regression matrix for depth of rutting.
No. Constants Regression Coefficients Statistics
AV sz Rec Pen-5 Gram?m ESAL“ R2 S.E

1 3.396 -0.315 0.462 0.955
2 5.295 -1.1 1 1 0.0656 0.716 0.698
3 2.827 -0.972 0.0580 0.269 0.806 0.582
4 2.657 -0.915 0.0549 0.253 0.0672 0.824 0.558
5 2.350 -0.830 0.0506 0.193 0.0936 0.102 0.852 0.516

Table 5.9 : Residual statistics of equation 5.4

Statistics Minimum Maximum Mean Std. Deviation N

Predicted Value 0.060 2.007 0.630 0.470 73

Std. Predicted Value -1.214 2.931 0.000 1.000 73

Standard Error Pred 0.080 0.296 0.141 0.045 73

Value

Adjusted Predicted 0.133 5.005 1.605 1.185 73

Value

Residual -1.543 1.246 0.000 0.498 73

Std. Residual -2.989 2.414 0.000 0.965 73

 

 

 

 

119

Further, the inaccuracy of the ESALS data can also be depicted fiom table 5.1. It can
be seen that the ESALS number varies substantially from one point to another on the same
road. For example sections 44A & 45A are both located on N-5 highway within 2 to 3 km
distance from each other. The ESALS for section 44A and 45A are 9.8 and 19.71

respectively.

5.4.5 Sensitivity Analysis and Engineering Interpretation

The statistical model presented by equation 5.4 was used to predict the average rut
depths over the selected sites. Figure 5.1 depicts the values of the average predicted
depths of rutting versus the average observed ones. The straight line in the Figure
represents the percent errors between the predicted and the observed depths of rutting. It
can be seen that the error in the predicted rut values is substantially for low rut &
relatively accurate with high rut values. Since high pavement rutting is typically
associated with higher ESALS, the error of the low rut values may indicate inaccuracy in
the ESALS data. Examination of figure 5.2 indicates that 2/3 of the predicted rut depths
are within 40 percent error of the observed values. This indicates that the statistical model
is relatively accurate. One additional point should be made herein is that, the rut depth of
a pavement structure is a fimction of several variables including the stiffness and
thickness of the various pavement layers, the properties of the materials in the AC mix,
the traffic volume, load, and environmental factors. In this study, only the asphalt mix

properties and the traffic load and volume data are used.

120

 

A33 :32— .3— 695 our—93‘

Average Measured Rut Depth (cm)

Average measured versus the predicted rut depth at 73 sites.

Figure 5.1

Frequecy

121

30

 

 

 

0-20 20-40 40-60 60-80 80- 100 >100

Percent Error Range

Figure 5.2: Percent difference between the average observed and the predicted
rut depths for 73 sites.

 

122

Consequently, one should not expect the statistical model presented in equation 5.4 to fiilly

explain the variations in the rut depths.

A sensitivity analysis of equation 5.4 was performed to determine the effect of each

variable in the equation on the predicted rut depth at the 73 core locations. The sensitivity of

the model to each variable was analyzed such that:

1.

2.

The values of three variables in the equation were held constant.

The values of the one constant variable were changed from a low, to a medium, to
a high value and values of other two variables were kept constant over the values,
within the range of values of that variable. Since most of the sites have zero
longitudinal grade, so its average was taken as zero.

The value of the fourth variable in question was varied from one end of its range to

the other.

Results of the sensitivity analysis are presented and discussed below.

Air Voids - Figure 5.3 depicts the sensitivity of the predicted rut depth to the percent air

voids for average values of grade (since most of the data points have zero grade, so average

is replaced by zero value) and Rec Pen (57) and three levels of ESALS (1.5, 8, 15).

Examination of the figure reveals that:

1.

Decreasing the percent air voids below three percent level causes significant
increase in the rut depth.
The rut depth increases as the percent air voids increases above eight percent level.

The rut depth increases with increasing the ESALS.

123

The result presented in figure 5 .3 and three observations stated above should be
viewed with extreme caution. The reason is that the air voids is highly correlated to the
percent asphalt content as it can be seen from table 5.3. This correlation or collinearity tends
to distort the true effect of the air voids on rut depth.

Field data suggests that the pavement constructed with air voids of less than three
percent, experiences excessive rutting. On the other hand, pavements constructed with 3 to
6 percent air voids show low rutting potential. Pavements constructed with air voids higher
than 6 percent will have higher rutting potential. The rut prediction model shows an
optimum air voids of 8 percent. This represents significant departure fiom the field
observations of optimum air voids of about 4.5 percent. Once again, the reason for such
distraction is the collinearity between the AC content and the air voids & visa versa. Since
AC contents cause an increase in the rut potential, the true effect of air voids on rutting

cannot be obtained from the given data.

Recovered Asphalt Penetration - Figure 5.4 depicts the sensitivity of the predicted rut
depth and the recovered asphalt penetration (Rec Pen) for average values of grade (zero)
and air voids (5.76) and three levels of ESALS. It can be seen that the rut depth increases
with an increase of recovered asphalt penetration. In view of this observation, lower grade
asphalt should be used to retard rutting. However, low grade asphalts have high temperature
cracking potential. Therefore, the selection of proper asphalt grade should be made afier
balancing the rutting and temperature cracking potentials. In this regard the new SHRP

specifications concerning AC grades are recommended to be used in Pakistan.

124

In addition, a significant collinearity between ESALS and asphalt penetration exists (see
table 5.3). Higher ESALS implies older pavements and longer oxidation time which causes
lower recovered asphalt penetrations. Unfortunately, such collinearity can be resolved but
by substantially increasing the sample size so that the data can be stratified. This is possible
if and only if, 9a complete pavement management system data bank exists. This not the case

in Pakistan.

ESALS - Figure 5.5 depicts the effects of ESALS on rut depth for average value of
Gradient( zero) and air voids (5.76) three levels of Rec Pen. It is evident from the figure
that rut depth increases with increasing number of 18 kip ESALS. The rate of increase
decreases with increasing ESALS. This signifies that rutting due to densification is
significant after opening the pavement to traffic. The rate of rutting decreases as the AC

layer is subjected to a higher degree of densification (compaction) due to traffic.

Gradient - Figure 5.6 depicts the sensitivity of the predicted rut depth to the
gradient(grade) of the road section with average values of Rec Pen (57 and air voids (5.76)
and three levels of ESALS.. The rut depth increases slightly in the initial stage of the
gradient and then rises sharply and substantially after 1.5 percent of grade. These
observations were expected because higher grades retard the speed of vehicles. Lower speed
induces longer loading time and higher rut. Excessive overloading is the normal practice in
Pakistan. The effect of overloading is enhanced many fold, in areas where roads also have a

longitudinal grade.

125

5.00

 

 

 

4,5011%.-7 . . K A “... -A . A .

 

  

 

ESAL=1.S, Rec Pen=57, Grade=0
------ ESAL=8, Rec Pen=57, Grade==0
ESAL-=15, Rec Pen= 57, Grade==0

 

 

3.50 ,

 

 

 

 

1

3.00 4

2.50 ,

 

Rut Depth (cm)

2.00 <

 

1.507

 

 

 

1.00 7 {— is

0.50 ,A

 

 

 

 

 

1 e- A

 

 

 

0.00 i. i
2
Percent Air Voids

Figure 5.3: Rut depth as a function of air voids for average values of Rec Pen,
Grade and three levels of ESALS (millions).

14

5.00

4.50 »

4.00 ‘-

3.50 .0

3.00 4

Rut Depth (cm)

2.00 <

1.501

1.00‘

0.50 <

0.00

126

 

2.50 «

 

 

 

L311111

 

 

ESA1F1.5, AV=S.76, Grade=0 rm

------ ESAL-=8, AV=S.76, Grade=0
ESAL=15, AV= 5.76, Grade=0

 

 

 

 

1111

 

 

50 60 70 80 90 1 00 1 10 1 20

Recovered Asphalt Penetration

Figure 5.4: Rut depth as a function of Rec Pen for average values of air voids,

Grade and three levels of ESALS(millions).

5.00

4.50 ~

4.00

3.50 ~

Rut Depth (cm)

Figure 5.5: Rut depth as a function of ESALs for average values of air voids,

127

 

 

 

 

 

l
. l -

1 . 1. 1
1 1 1

1

HI
111

1

i

 

 

 

 

Rec Pen=40, AV=S.76, Grade=0
------ Rec Pen=60, AV=S.76, Grade=0
Rec Pen=90, AV= 5.76, Grade=0

 

 

 

 

 

. A.“ .
1

 

 

 

 

 

 

 

 

 

6 a 10 {2 14 1e
ESALs (Millions)

Grade and three levels of Rec Pen.

18

20

22

24

26

128

5.00

 

 

 

 

 

 

 

 

 

ESAL=1.5, Rec Pen=57, AV=S.76

_ . _ _ ------ ESAL-'8, Rec Pen=57, AV=S.76

 

4.50 A 1

 

ESAIFIS, Rec Pen=57, AV= 5.76

 

 

 

 

4.00 ‘

 

3.50 ~

 

 

3.00 .

 

2.50 ‘

 

Rut Depth (cm)

2.00 -

 

 

 

 

 

1.50~

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

1.00«
0,50 1 1 1 __ 1 - A
1 1 1 1 1
. 1 L 1 .
o 0.5 1 1.5 2 2.5 3 3.5 4 4.5 5
Grade (Percent)

Figure 5.6: Rut depth as a function of grade for average values of Rec pen, air voids
and three levels of ESALS.

CHAPTER 6

SUMMARY, CONCLUSION AND RECOMMENDATIONS

6.1 SUMMARY

The effects of the asphalt mix variables on pavement rutting were investigated in
this study. Full depths AC cores were obtained from selected pavement sections. The
properties of the AC mixes were determined as described in chapter 4.

A rut prediction model was developed which is based on the asphalt mix
properties, cumulative 18 kip ESALs and gradient of the road sections. Sensitivity
analysis of the model was conducted to determine the behavior and influence of each
variable on pavement rutting. It is shown that the most significant variables affecting
pavement rutting are air voids, recovered asphalt penetration, cumulative ESALs and

gradient of the road sections.

6.2 CONCLUSIONS

Based on the test data, the analysis and literature on the subject, the following
conclusions were drawn:
1. Rut resistance of the AC mixes can be maximized by specifying a maximum of

3 to 5 percent air voids.

129

6.3

130

Asphalt viscosity/penetration has shown definite impact on rutting. Softer asphalt
grade are most susceptible to rutting.

The number of cumulative 18 kip ESALs effect pavement rutting. Sections
experiencing higher volumes of traffic/loads in a short span of service life have
shown extensive amount of rutting.

Gradient of the road sections (longitudinal slope) causes higher rut. Road grade
impedes vehicles speed which causes longer loading time.

Significant variations in the thickness of the AC layers and AC mix properties (air

voids etc.) were observed within one pavement section.

RECOMMENDATIONS

Based on the findings and conclusions of the study, the following

recommendations are made:

1.

To improve the rut resistance of AC mix, the air voids percentage between 3 to 5
percent be specified.

Measures be taken to rationalize and enforce the axle load limit to avoid excessive
damage to roads.

Gradient specification or grade be reduced if economically feasible.

Quality assurance/quality control practices regarding pavement construction and
AC mix manufacturing operations should be modified and enforced as lot of
variation in pavement thicknesses and mix properties were found within the same

section of road.

131

Fresh axle load and traffic volume surveys be undertaken to determine the
accurate and updated records of truck factor, growth rates and traffic counts,
which are required in pavement design. It will be more appropriate if weigh-in-
motion devices are installed to monitor traffic volume and loads. Traffic record
for newly constructed/rehabilitated roads should be maintained and pavement
performance be monitored regularly.

The rut model presented in this thesis be expanded to include the other
engineering properties of the AC mixes (e.g. elastic and plastic) and the properties

of the other pavement layers and subgrade.

10.

11.

12.

REFERENCES

Yoder EJ. and M. W. Witczak, "Principles of pavement design".A Wiley-
Interscience Publication, John Wiley and Sons, Inc,.1975.

American Association of State, Highway, and Transportation Officials. "AASHTO
Guide for design of pavement structures". Washington, D. C., 1986.

Jack Morris. "The prediction of permanent deformation and structural design
method for flexible pavements". Department of civil Engineering, University of
waterloo, Waterloo Ontario Canada.

Mukhtar H., "Reduction of pavement rutting and fatigue cracking". P.h.d Thesis
1993, Department of civil Engineering, Michigan State University USA.

Baladi G. Y.," "Integrated material and structural design method for flexible
pavements," Vol. 1: Technical Report, FHWA, 1988.

Majidzadeh Kamran and Measorvic S, “A constituative model for asphalt
concrete, application to rutting.” Paper # 929386, 71th meeting TRB 1992.

Report 5, Highway Research Board Special Report 61E, " The ASSHO Road Test,"
1962.

G. A. Huber and G. H. Heiman. “Effect of asphalt concrete parameter on rutting
performance. A field investigation”. AAPT. VOL. 56, 1987.

Jack Morris. Ralph C. G. Hass, Park Reilly, and Edward T. Hignell, "Permanent
deformation in asphalt pavement can be predicted". APT. 1974, VOL. 43.

Parker F and Brown E R, “Effects of aggregate properties on flexible pavements
rutting in ALABAMA.” Paper # 910760, 70th meeting TRB 1991.

Harold Von Quintus and Thomas W. Kennedy, “AAMAS Mixture properties
related to pavement performance”. APT. Vol 58 1989.

Lee , K. W.and Al-Dhalaan, M. A., “Rutting, Asphalt Mix-Design, and Proposed
Test Road in Saudi Arabia”, Implications of Aggregate in the Design,
Construction, and Performance of Flexible Pavements, ASTM STP 1016, H. G.
Schreuders and C. R. Marek, Eds, American Society for Testing Materials,
Philadelphia, 1989.

132

13.

14.

15.

16.

17.

18.

19.

20.

21.

22.

23.

24.

25.

133

Hughes C. S. and G. W. Maupin, Jr., "Experimental bituminous mixes to minimize
pavement rutting," APT Vol. 56, 1987.

Hsien H, Chen, Kurt M. Marshek, and Chhotel L, Saraf, "Effect of truck tire contact
pressure distribution on the design of flexible pavement: A three dimensional finite

element approach," Transportation Research Record (TRR), Transportation
Research Board (TRB), 1095, 1986.

Baqur S. A." Analysis of Tire Pressures in Pakistan.” MS Thesis Military College
of Engineering Risalpur, 1993.

Bonquist R. R . "Effects of tire pressure on flexible pavement response and
performance," Draft Report, Federal Highway Administration, Washington, D. C.,
1989.

Harry A. Smith, "Truck tire characteristics and asphalt concrete rutting," Paper No.
910035, Transportation Research Board (TRB) 70TH Annual Meeting, 1991.

Richard. D. Barksdale, "Practical application of fatigue and rutting tests on
bituminous based mixes". APT VOL.47,1978.

L.E Santucci and R. J. Schmidt, “The effects of asphalt properties on the fatigue
resistance of asphalt paving mixtures”. Asphalt Paving Technology VOL. 38, 1969.

Miller C. Ford, Jr. Pavement densification related to asphalt mix characteristics".
TRR.1178, 1988.

Jon a Epps and Carl L. Monismith," Influence of mix variables on flexural fatigue
properties of AC. “APT VOL. 38.

E. R. Brown, “Mix design and construction of asphalt concrete to support high tire
pressures”. Presented at AASHTO/FHWA Symposium on high pressure truck tires,
Austin, Texas Feb. 1987.

Carpenter H. Samuel, “ Permanent deformation: Field Evaluation” Paper # 930768,
Transportation Research Board ,72nd meeting, 1993.

Lister W. N and Addis R. R, “Field observations of rutting and their practical
implications” ,Transport and Research Laboratory, UK. Department of the
Environment Crowthome, England.

“Essential physical properties of Sand, Gravel, Slag and Broken Stone for use in
Bituminous Pavements”. Better Roads and Streets. VOL. 6, No. 3.

26.

27.

28.

29.

30.

31.

32.

33.

34.

35.

36.

37.

38.

39.

134

“Mix design methods for asphalt concrete”. Asphalt institute manual series NO.
2(MS-2).

Hicks, R.G. and David C. ESCH, “State of the Art on rutting in asphalt concrete”.
Alaska DOT and PF Fairbanks, AK 99701.

E R Brown and Charlas E. Bassett, “ Effects of maximum aggregate size on rutting
potential and other properties of asphalt aggregate mixtures”. TRR. 1259, 1990.

Kamyar Mahboub and David L. Allen, “Characterization of rutting potential of
large stone asphalt mixes in Kentucky “. TR 1259 1990.

Moreland Herrin and W.H. Goetz, “Effect of aggregate shape on stability of
bituminous mixes”. Joint Highways Research Projects, Purdue University. HRB.
Proceedings 33rd Annual Meeting 1954.

Mohammed Osamn Khalifa and Moreland Herrin, “The behavior of asphaltic
concrete constructed with large sized aggregate”. APT Vol.39.

Kandhal P S, Stephen A Cross and E R Brown, “Heavy duty asphalt pavements in
Pennsylavinia: Evaluation for rutting.”

W.S. Mendenhall, Jr., Chairman, “Ad hoc task force on asphalt pavement rutting
and stripping”. August 14, 1987.

Joe W. Button, Dario Perdomo and Robert L. Lytton, “ Influence of aggregate on
rutting in asphalt concrete pavements”. TRR 1259 1990.

M.A. Young and G. Y. Baladi, “Repeated load triaxial testing, state of the art”.
Michigan State University. Department of Civil Engineering.

Ronald E. Reese and Joseph L. California Desert Test Road - A step closer to

performance based specifications published in Asphalt paving technology Vol 62,
1993.

Neil C. Kruts, Raj Siddharthan and Mary Stroup-Gradiner, “Investigation of
rutting potential using static creep testing on polymer modified AC mixture”.
Paper # 91-0705, 70th meeting TRB 1991.

E. R. Brown and Stephen A. Cross, “A study of in-place rutting of asphalt
pavements”. APT. VOL.58 1989.

M.J. Hensley and Rita B. Leahy, “Asphalt concrete mixtures as related to
pavement rutting: Case study”. TRR. 1217 1989.

40.

41.

42.

43.

44.

44.

46.

47.

48.

49.

50.

51.

52.

53.

135

Dale S. Decker and Joseph L. Goodrich, “ Asphalt cement properties related to
pavement performance”. APT VOL. 58 1989.

Collins J H, Bouldin M G Gells R and Berker A, “ Improved performance of
paving asphalt by Polymer modifications”.

RD. Barksdale, “Compressive stress pulse time in flexible pavements for use in
dynamic testing”. Highways Research Board (HRB). Highways Research Record
(HRR) NO. 345, 1971.

RD. Barksdale, “Laboratory Evaluation of rutting in base course material”. Pro. 3rd
International Conference on the structural design of asphalt pavements, London
1972.

W. Heukelon and A.J.G. Klomp, “Consideration of calculated strain at various
depths in connection with the stability of asphalt pavements”. Pro. 2nd International
Conference on the structural design of asphalt pavement, Ann Arbor 1967.

J .E. Romain, “Rut depth prediction in asphalt pavement” Pro. 3rd International
Conference on the structural design of asphalt pavement, London 1972.

Richard D. Barksdale and Gerald A.Lenards, “Predicting performance of
bituminous surfaced pavements”. Pro. 2nd International Conference on the
structural design of asphalt pavement, Ann Arbor 1967.

Soussou, J .E and Moavenzadeh, F. “Statistical Characteristics of fatigue damage
accumulation in flexible pavements”. American Society of Testing and Material
special technical papers No. 561 1973.

Rita B. Leahy and Matthew W. Witczak. “The influence of test conditions and

asphalt concrete mix parameters on permanent deformation co-efficient Alfa and
Mu,”. APT Vol 60, 1991

Baladi G. Y., “Fatigue life and permanent deformation characteristics of asphalt
concrete mixes” , TRR, TRB, 1227, 1989.

Cardoso S H and Witczak WM,“Pavement deformation for flexible pavements.”

Matthews J M and Pandey B B, “ Performance of flexible pavements.”
Traffic Volume Data 1995 by NTRC Islamabad.

Traffic count NWFP 1992.

54.

55.

56.

57.

58.

59.

60.

61.

62.

63.

64.

65.

66.

67.

136

Traffic Survey Punjab 1992.

ALMEC Corporation Pacific Consultants International, “The Study on National
Transportation Plan in the Islamic Republic of Pakistan”. Final Report Volume II,
NTRC (Growth Factors), February 1995.

Iqbal Zafar, “Axle load survey to establish latest truck factors between Taxila and
Lahore” MS Thesis 1993.

Axle load study on National Highways 1995 by NTRC Islamabad.

Axle load studies in Punjab by Road research and material testing institute
Lahore.

Tresca, H “Memoire Sur L’ ecoulement Des Corps Solides, “Mem. Pres. Par div.
Savants 18, 1868.

Venant B. de Saint, “Memoire sur L’establishment des Equantions Differentiales
des Mouvements Interiors opers Dans lrd Corps Solids Ductiles au Dela des Limits
au L’elasticite Pourrait les Ramener a Leur Premier Etat, “CR. Acad. Sci. (Paris),
70,1870.\

Levy M., “Memories sue L’equations Generales des Mouvements interieurs des
Corps Solides Ductiles au Dela des Limits on L’elasticite Pourrait les Ramener a
Leur Premier Etat,” CR. Acad. Sci. (Paris), 70, 1870.

W. Prager, “An Introduction to Plasticity,” Addison-Wesley Publishing Company,
Inc., London, England. 1959.

ERES Consultants, Inc.”Pavement design principles and practices” Charnpaign
Illinois, National Highway Institute, Washington, DC, 1987.

Majidzadeh, U. and G. Ilves.Flexible pavement overlay design procedures, volume
1, Evaluation and modification of the design methods” Final report FHWA-RD-
82a32. 1981.

John O. Rawlings, “Applied regression analysis, a research tool”. Wadsworthand
Brooks/Cole, Statistics/Probability Series 1989.

David A. Belsley, Edwin Kuh, Roy E. Welsch, “Regression diagnostics, identifying
influential data and sources of collinearity”. Wiley series in probability and

mathematical statistics.

Marija J. Noruses/SPSS Inc, “SPSS statistical data analysis”.

 

MICHIGAN STATE UNIV. LIBRARIES
1111111111111111111111111111111111111111111111111111111111111111
31293015706116