EFFECT OF SUBGRADE CONDITIONS ON PAVEMENT ANALYSIS AND

PERFORMANCE PREDICTION: A STUDY FOR IDAHO CONDITIONS

by

Md Jibon

A thesis

submitted in partial fulfilment

of the requirements for the degree of

Master of Science in Civil Engineering

Boise State University

August 2019
 © 2019

Md. Jibon

ALL RIGHTS RESERVED

 BOISE STATE UNIVERSITY GRADUATE COLLEGE

DEFENSE COMMITTEE AND FINAL READING APPROVALS

of the thesis submitted by

Md Jibon

Thesis Title: Effect of Subgrade Conditions on Pavement Analysis and Performance

Prediction: A Study for Idaho Conditions

Date of Final Oral Examination: 2 August 2019

The following individuals read and discussed the thesis submitted by student Md. Jibon,
and they evaluated his presentation and response to questions during the final oral
examination. They found that the student passed the final oral examination.

Debakanta Mishra, Ph.D. Chair, Supervisory Committee

Bhaskar Chittoori, Ph.D. Member, Supervisory Committee

Emad Kassem, Ph.D. Member, Supervisory Committee

The final reading approval of the thesis was granted by Debakanta Mishra, Ph.D., Chair
of the Supervisory Committee. The thesis was approved by the Graduate College.
 DEDICATION

Dedicated to my parents for their unconditional love and support.

iv
 ACKNOWLEDGEMENTS

I would like to express my deepest appreciation to my supervisor Dr. Debakanta

Mishra for his guidance, constant supervision and support toward this journey of

completing a master’s degree in civil engineering. This research would not be completed

without his valuable instructions and encouragement. I would also like to thank my

committee members Dr. Bhaskar Chittoori and Dr. Emad Kassem for their guidance and

for being a part of my supervisory committee.

I want to express my gratitude to all faculty members in the Civil Engineering

Department at Boise State University for always encouraging students to achieve their

highest potential. I also sincerely appreciate the support of my family members, friends

and all graduate students in the Civil Engineering Department at Boise State University.

Finally, I would like to acknowledge the support from several engineers in the

Idaho Transportation Department (ITD), Especially Mr. Mike Santi of ITD Headquarters,

and Mr. John Arambarri of ITD District 3. Collecting all the soil specimens for laboratory

testing would not have been possible without their extensive support.

v
 ABSTRACT

The Mechanistic-Empirical (M-E) pavement design approach detailed in the

Mechanistic-Empirical Pavement Design Guide (MEPDG), and subsequently implemented

through AASHTOWare Pavement ME Design relies extensively on detailed material

properties that ultimately govern the analysis and performance prediction results. For

unbound materials like soils and aggregates, Resilient Modulus (MR) is the most critical

input parameter affecting layer response under vehicular and environmental loading.

Representing a material’s ability to ‘recover’ after loading, resilient modulus is determined

in the laboratory through repeated load triaxial testing. Although the original test protocol

to measure the resilient modulus value of a soil or aggregate was developed back in the

1980’s, this test is still not widely used by state highway agencies because it is

cumbersome, and requires significant investments towards equipment and personnel

training. Accordingly, most agencies rely on correlation equations to predict the resilient

modulus values for soils and aggregates from other easy-to-determine material properties.

However, these correlation equations are mostly region specific, and therefore, do not

produce adequate results across different geographic regions. This has led several state

highway agencies to undertake local calibration efforts for improved prediction of material

properties.

Over the past decade, the Idaho Transportation Department (ITD) has invested

significant resources to facilitate state-wide implementation of mechanistic-empirical

pavement design practices. A research study was recently undertaken by ITD to develop a

vi
database of resilient modulus properties for different soils and aggregates commonly used

in the state of Idaho for pavement applications. Another objective of the study was to assess

the adequacy of different correlation equations currently available to predict soil and

aggregate resilient modulus from easy-to-determine material (strength and index)

properties. This Master’s thesis is based on tasks carried out under the scope of the above-

mentioned project, and focuses on laboratory characterization and analysis of

representative subgrade soil types collected from across Idaho. An extensive laboratory

test matrix was developed involving commonly used mechanical and index tests, repeated

load triaxial tests for resilient modulus determination, as well as tests to study the soil

permanent deformation (plastic strain) behavior. Effect of moisture variation on soil

strength, modulus, and permanent deformation properties was also studied by testing soil

specimens at three different moisture contents. The test results were thoroughly analyzed

to evaluate the feasibility of predicting resilient modulus from other material properties.

Findings from this research effort have been documented in the form of two journal

manuscripts. The first manuscript highlights the importance of using adequate subgrade

resilient modulus values during pavement design. Eight different soil types were randomly

selected from a total of sixteen soil types, and the corresponding laboratory test results

were used to highlight the limitations of ITD’s current approach with assumed resilient

modulus values. The second manuscript focuses on highlighting the importance of

unbound material permanent deformation characterization during pavement design, and

how small changes in moisture content can lead to significant differences in the rutting

behavior of subgrade soils. First, a new permanent deformation testing protocol was

developed to simulate typical stress states experienced by subgrade layers under vehicular

vii
loading. Subsequently, permanent deformation tests were carried out on subgrade soil types

collected from two distinctly different regions in Idaho as far as annual precipitation is

concerned. Tests were conducted at three different moisture contents to highlight how the

rutting potential of the subgrade may change significantly based on site precipitation and

drainage characteristics. Finally, recommendations were made regarding how state

highway agencies can accurately represent resilient modulus properties of soils during

pavement analysis and performance prediction using AASHTOWare® Pavement ME

Design.

viii
 TABLE OF CONTENTS

DEDICATION ...............................................................................................................iv

ACKNOWLEDGEMENTS ............................................................................................. v

ABSTRACT ...................................................................................................................vi

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

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

CHAPTER 1: INTRODUCTION AND BACKGROUND ............................................... 1

Background and Problem Statement ..................................................................... 1

Objectives and Research Tasks ............................................................................ 2

Research Question and Hypothesis ....................................................................... 3

Resilient Behavior of Unbound Materials – A Brief Overview .................. 4

Organization of the Thesis ................................................................................... 7

References ........................................................................................................... 8

CHAPTER 2: MANUSCRIPT ONE - LABORATORY CHARACTERIZATION OF

FINE-GRAINED SOILS FOR PAVEMENT ME DESIGN IMPLEMENTATION IN
IDAHO.......................................................................................................................... 10

Abstract ............................................................................................................. 10

Introduction ....................................................................................................... 11

Factors Affecting Resilient Response of Fine-Grained Soils ............................... 12

Pavement ME Design Implementation and Need for Correlation Development .. 13

Objective of Study ............................................................................................. 14

ix
 Materials and Methods ....................................................................................... 15

Grain Size Distribution ........................................................................... 15

Compaction Characteristics and Soil Classification ................................ 17

Unsoaked California Bearing Ratio (CBR) Testing ................................. 18

Resilient Modulus Testing ...................................................................... 19

Unconfined Compressive Strength.......................................................... 21

Analysis of Test Results ..................................................................................... 22

Resilient Modulus Model Parameters ..................................................... 22

Effect of Moisture Content and Deviatoric Stress on Soil Resilient

Modulus ................................................................................................. 25

Predicting Soil Resilient Modulus from Index Properties ................................... 29

Specifying One Resilient Modulus Value for an Unbound Material ........ 29

Correlation Development to Predict Summary Resilient Modulus

Values .................................................................................................... 31

Effect of Soil Resilient Modulus on Predicted Pavement Performance Using

AASHTOWare Pavement ME Design .............................................................. 36

Effect of Soil Resilient Modulus on Critical Pavement Response Parameters ..... 40

Use of Simple Damage Models to Compare Pavement Life .................... 43

Effect of Moisture-Induced Subgrade Resilient Modulus Variation on Pavement

Performance Prediction ...................................................................................... 44

Effect of Soil Moisture on Critical Pavement Response Parameters – Results from

Finite Element Analysis ..................................................................................... 47

Summary and Conclusions ................................................................................. 49

References ......................................................................................................... 51

CHAPTER 3: MANUSCRIPT TWO – MOISTURE SENSITIVITY OF TYPICAL

IDAHO SUBGRADE SOILS AND THEIR IMPLICATION ON M-E PAVEMENT
DESIGN ........................................................................................................................ 54

x
 Abstract ............................................................................................................. 54

Introduction ....................................................................................................... 56

Material Selection .............................................................................................. 59

Preliminary Soil Characterization and Shear Strength Measurements ................. 61

Grain Size Distribution ........................................................................... 61

Moisture Density Relationship ................................................................ 62

Strength Properties of Soils for Typical Pavement Design ...................... 63

Resilient Response for Pavement ME Design ..................................................... 66

Finite Element Analysis of Typical Pavement Sections ...................................... 69

Laboratory Tests to Characterize the Permanent Deformation Potential of Soils . 72

PD test program...................................................................................... 72

Analysis of PD Test Results ................................................................... 73

Summary and Conclusions ................................................................................. 76

Limitations of Current Study and Recommendations for Future Research .......... 77

References ......................................................................................................... 78

CHAPTER 4: SUMMARY AND CONCLUSIONS ...................................................... 80

Summary ........................................................................................................... 80

Conclusions ....................................................................................................... 81

xi
 LIST OF TABLES

Table 2-1 Percentage Gravel, Sand, Silt and Clay Content in Different Types of Soils
...............................................................................................................16

Table 2-2 Soil Index Properties and Compaction Characteristics ............................18

Table 2-3 Unsoaked CBR Values for the Eight Soil Materials at Different Moisture
Contents .................................................................................................19

Table 2-4 Resilient Modulus of Soils at Different Stress States Tested at OMC ...... 21

Table 2-5 Peak Compressive Strength of Soils at Different Moisture Contents ....... 22

Table 2-6 MEPDG Model Parameters for the Eight Soil Materials Corresponding to
Three Different Moisture Contents .........................................................24

Table 2-7 Multiple Regression Analysis Output for the Five Models ......................35

Table 2-8 Comparison of Pavement Performances Obtained from Pavement ME. .. 39

Table 2-9 Sensitivity of Subgrade Resilient Modulus in Damage Predictions ......... 44

Table 2-10 Required Layer Thickness of Pavement Built in Tested Subgrade ........... 45

Table 2-11 Predicted Pavement Responses Found from ILLIPAVE Stress

Dependency Tool ...................................................................................48

Table 3-1 Soil Index Properties and Classification ..................................................62

Table 3-2 Moisture Contents Corresponding to Dry, OMC, and Wet Conditions for
the Two Soil Types .................................................................................63

Table 3-3 California Bearing Ratio and Unconfined Compressive Strength of Soils64

Table 3-4 MEPDG and Uzan Model Parameters of Soils at Different Moisture
Conditions ..............................................................................................69

Table 3-5 Summary of Material Properties for ILLI-PAVE Analysis ...................... 70

xii
 LIST OF FIGURES

Figure 1-1 Schematic of moving wheel loads induced stress state in pavement
material (Lekarp et al., 2000) ....................................................................5

Figure 1-2 Stress strain response of unbound material under one cycle of loading
(Mishra, 2012) ..........................................................................................6

Figure 1-3 Graphical representation for defining Resilient Modulus of unbound

materials (Tutumluer, 2013) .....................................................................7

Figure 2-1 Grain size distribution of different soil types ..........................................16

Figure 2-2 Variation of MR with different moisture content for plastic soils .............26

Figure 2-3 Variation of MR with different moisture content for non-plastic soils ......28

Figure 2-4 Distribution of summary modulus data tested at each moisture content ...32

Figure 2-5 Minimum required layer thicknesses while using all ITD-recommended
values as material inputs. ........................................................................38

Figure 2-6 Predicted rutting for (a) recommended subgrade (b) tested subgrade ....... 40

Figure 2-7 Variation in predicted pavement performances; a) vertical displacement b)

vertical stress and c) horizontal strain with depth starting from top of
pavement. ...............................................................................................42

Figure 2-8 Effect of subgrade moisture on a) IRI and b) rut depth of flexible
pavement ................................................................................................46

Figure 3-1 Annual average precipitation map for Idaho ...........................................60

Figure 3-2 Grain size distribution of selected two soil types .....................................61

Figure 3-3 Compaction moisture-density characteristics for the two soils established
following the AASHTO T 99 procedure .................................................63

Figure 3-4 (a) California Bearing Ratio and (b) Unconfined Compressive Strength of
soils at different conditions .....................................................................65

xiii
Figure 3-5 Resilient Modulus values for the two soils under different moisture
conditions: (a) soil type S1 and (b) soil type S2 ........................................67

Figure 3-6 Tensile strain at bottom of HMA layer (b) Compressive strain at top of
subgrade (c) Deviator stress at top of subgrade .......................................71

Figure 3-7 Accumulation of plastic strain in soil S1 when tested at two different
deviator stress values: (a) σ d = 41.4 kPa ; and (b) σ d = 55.2 kPa ........... 74

Figure 3-8 Accumulation of plastic strain in soil S2 when tested at (a) 41.4 kPa
deviator stress (b) 55.2 kPa deviator stress ..............................................75

xiv
 1

CHAPTER ONE: INTRODUCTION AND BACKGROUND

Background and Problem Statement

The subgrade constitutes the weakest layer in a pavement system, and can lead to

significant pavement damage if not protected from excessive stress levels during the

pavement service life. The design and construction of a well-performing, economical

pavement system requires thorough knowledge of the subgrade properties. Since the

release of its 1986 edition (AASHTO 1986), the American Association of State Highway

and Transportation Officials (AASHTO) pavement design method has considered

‘Resilient Modulus’ (MR) to be a key soil input during the pavement design process (Taylor

& Timm 2009). The recently released Mechanistic Empirical (M-E) Pavement Design

approach implemented through AASHTOWare® Pavement ME Design uses the soil

resilient modulus as a critical input parameter for analysis and performance predication

under vehicular as well as environmental loading. Resilient modulus properties of a soil

can be determined in the laboratory through repeated load triaxial testing using

specifications such as AASHTO T 307 (AASHTO T307 1999), or more recently, NCHRP

1-28A (Barksdale et al. 1997). However, these tests are quite complex in nature, and

require significant investments towards equipment procurement and personnel training.

This has led to several state highway agencies avoiding the performance of repeated load

triaxial tests, and using correlation equations to predict soil MR values from other easy-to-

measure mechanical and index properties. Although researchers have developed several

such correlation equations over the years (Robnett and Thompson 1973; Jones & Witczak
 2

1977; Elliott et al. 1988; Drumm et al. 1990; Woolstrum 1990 ), these equations are not

able to consistently predict MR from soil index properties (Mokwa & Akin 2009).

Accordingly, MR values predicted using such correlation equations can often under-predict

or over-predict the soil modulus, thus leading to erroneous pavement design. Certain

studies in the past have shown that the correlation equations work with reasonable accuracy

on a regional basis. In other words, certain correlation equations may work fairly well for

soils from a particular region, while not working well when applied to soils from a different

region. This has led several state highway agencies to undertake regional correlation

development efforts to predict soil resilient modulus from simple mechanical and/or index

properties.

Objectives and Research Tasks

Over the past decade, the Idaho Transportation Department (ITD) has committed

significant resources towards facilitating statewide implementation of M-E pavement

design practices. The design method currently used by ITD relies on the soil Resistance

Value (R-value) to calculate required pavement thicknesses; this method has been known

to result in significant over designs. As a part of ITD’s current M-E implementation efforts,

a research study was recently initiated to characterize typical soils and aggregates

commonly used in Idaho for resilient modulus characteristics. The objective was to build

a database of test results that would subsequently be used by ITD engineers during M-E

pavement design. Moreover, the project also aimed at assessing the accuracy of different

correlation equations developed to predict the resilient modulus values for soils and

aggregates from easy-to-measure mechanical and index properties.

 3

The research work reported in the current master’s thesis was based on tasks

performed under the scope of the above-mentioned ITD-sponsored research project, and

focused primarily on the subgrade soils collected from different parts of Idaho. Besides

testing the soils for database development, this master’s thesis aimed at highlighting

shortcomings associated with ITD’s current practice of assuming standardized resilient

modulus values based on soil types (based on the Unified or AASHTO soil classification

systems). Several example pavement sections were analyzed using AASHTOWare®

Pavement ME Design to highlight how the predicted pavement performance is affected by

varying resilient modulus values. Soil specimens were tested in the laboratory under three

different moisture contents, and the corresponding resilient modulus values were used

during the pavement analysis and performance prediction step to highlight the importance

of testing the soils at accurate moisture contents.

Finally, this master’s thesis effort also focused on studying the permanent

deformation (plastic strain) characteristics of the soils under repeated loading. A new

permanent deformation testing protocol was developed, and test results were used to

highlight why it is important to study both the resilient and plastic strain properties of

subgrade soils for design and construction of economical, well-performing flexible

pavement sections.

Research Question and Hypothesis

The research performed towards completion of the current master’s thesis was aimed

at answering the following research questions:

1. Will results from pavement analysis and performance prediction using

AASHTOWare® Pavement ME Design change significantly if assumed

 4

standardized soil properties are replaced by those specifically determined in the

laboratory?

2. Can a simplified permanent deformation testing protocol be developed in the

laboratory to correlate with rutting patterns predicted using AASHTOWare®

Pavement ME Design?

The overarching research hypothesis for this master’s thesis was: “better pavement

performance prediction can be achieved by using locally calibrated soil properties”.

Details of the tasks carried out to answer the above-listed research questions, and

the corresponding results have been reported in this master’s thesis in the form of two

independent manuscripts that constitute chapters 2 and 3 of this thesis. As the last

subsection of the current chapter, a brief overview of resilient behavior of unbound

materials under loading has been presented. This discussion will present the background

necessary for the two manuscripts included as chapters 2 and 3.

Resilient Behavior of Unbound Materials – A Brief Overview

A detailed understanding of unbound material behavior under loading is required

to facilitate the implementation of mechanistic-empirical pavement design practices.

Inadequate design and construction of unbound layers such as aggregate base/subbase

layers as well as the subgrade layer can adversely affect pavement performance. Pavement

materials experience a stress pulse which has normal, horizontal and shear stress

components due to moving nature of the wheel load (Lekarp et al. 2000). Figure 1-1 shows

the stress states applied to a typical soil element within the pavement structure upon the

passage of a wheel load (Lekarp et al. 2000). The unbound layers experience positive

vertical and horizontal stresses whereas the shear stress turns negative from positive as the
 5

load passes over the element. This reversal nature of shear stress leads to complex load

deformation behavior because of principal axis rotation.

Figure 1-1 Schematic of moving wheel loads induced stress state in pavement
material (Lekarp et al., 2000)

Unbound materials subjected to traffic-induced stress pulses experience both elastic

and plastic deformations which are commonly known as ‘resilient’ and ‘permanent’

deformations in pavement applications, respectively. The response of a typical unbound

material after being subjected to one cycle of pulse loading is shown in Figure 1-2 (Mishra

2012). These resilient and plastic components of total strain depend on several factors such

as number and magnitude of traffic load applications, speed of traffic operation, thickness

of undelaying and overlaying layers, quality of materials etc. The accumulation of plastic

strain gradually decreases with the number of load applications for typical unbound

materials during the pavement service life. Generally, well-compacted unbound layers
 6

become stable during the construction phase, and all subsequent loadings lead to

deformations that are elastic in nature.

Figure 1-2 Stress strain response of unbound material under one cycle of loading
(Mishra, 2012)

Ideally, a well-constructed pavement layer does not accumulate any permanent

deformation during repeated traffic loadings which is a key assumption inherent to

mechanistic empirical pavement design. Pavement ME considers only elastic response of

materials for predicting critical pavement response parameters. Resilient Modulus is a

fundamental material property which is defined as the ratio of peak deviatoric stress to

recoverable strain under repeated loading. Figure 1-3 shows a typical stress-strain plot to

define the concept of resilient modulus. Soil characterization under repeated loading helps

to understand material behavior under different stress fields, and facilitates accurate,

mechanics-based pavement analysis for pavement response calculation.

 7

Figure 1-3 Graphical representation for defining Resilient Modulus of unbound

materials (Tutumluer, 2013)

Subgrade soils are stress-dependent materials, and the resilient modulus value of a

soil depends on the applied stress states. Resilient modulus characterization of soils will

help understand how the subgrade layer will behave due to traffic-induced stress pulses. It

is also necessary to understand how moisture in subgrade affects resilient properties of

soils and the corresponding pavement performances. In the absence of detailed laboratory

testing data, state highway agencies often adopt national-level default values to represent

the subgrade soil properties during M-E pavement design. Such assumptions may lead to

design of pavement structures that are inadequate to sustain the design loads. Considering

these issues, a research effort was undertaken as a part of this master’s thesis work to

accurately characterize subgrade soils in the laboratory to facilitate the implementation of

M-E pavement design practices.

Organization of the Thesis

This Master’s thesis document contains four chapters. Chapter 2 includes all

findings from the first manuscript. The title of the manuscript is, “Laboratory
 8

Characterization of Fine-Grained Soils for Pavement ME Design Implementation in

Idaho”. Chapter 3 comprises findings reported in second manuscript, titled “Moisture

Sensitivity of Typical Idaho Subgrade Soils and their Implication on ME Pavement

Design”. Summary of results, conclusion and recommendations for future research have

been presented in Chapter- 4.

References

AASHTO Guide for Design of Pavement Structures. American Association of State

Highway and Transportation Officials, Wnshington, D.C., 1986.

AASHTO T307. Standard Method of Test for Determining the Resilient Modulus of
Soils and Aggregate Materials. AASHTO’S Standard Specification for
Transportation Material and Methods of Sample and Testing, 1999.

Barksdale, R. D., J. Alba, N. P. Khosla, R. Kim, P. C. Lambe, & M. S.

Rahman. Laboratory Determination of Resilient Modulus for Flexible Pavement
Design. Publication NCHRP Project 1-28. NCHRP, Transportation Research
Board, Washington D.C., 1997.

Drumm, E.C., Y. Boateng-Poku, and T. J. Pierce. Estimation of Subgrade Resilient

Modulus from Standard Tests. Journal of Geotechnical Engineering, 116(5),
1990, pp. 774-789.

Elliott, R. P., S. I. Thornton, K. Y. Foo, K. W. Siew, and R. Woodbridge. Resilient

Properties of Arkansas Subgrades. Final report, No. FHWA/AR-89/004, 1988
https://trid.trb.org/view.aspx?id=297506.

Jones, M. P. and M. W. Witczak. Subgrade Modulus on the San Diego Test Road. In
Transportation Research Record 641, Transportation Research Board, National
Research Council, Washington, DC, 1977, pp. 1-6.
https://trid.trb.org/view.aspx?id=73467

Lekarp, F., U. Isacsson, & A. Dawson. “State of the art. I: Resilient response of unbound
aggregates”. Journal of Transportation Engineering, 126(1), 66-75. 2000.
 9

Mokwa, R. L., & M. Akin. Measurement and Evaluation of Subgrade Soil Parameters:
Phase I-synthesis of literature. Publication FHWA/MT-09-006/8199. FHWA,
U.S. Department of Transportation, 2009.

Mishra, D. Aggregate Characteristics Affecting Response and Performance of

Unsurfaced Pavements on Weak Subgrades. Ph.D. Dissertation, University of
Illinois at Urbana-Champaign. 2012.

Robnett, Q. L., and M. R. Thompson. Resilient Properties of Subgrade Soils, Phase I –

Development of Testing Procedure. Illinois Department of Transportation, Interim
Report,1973, pp. 45. https://apps.ict.illinois.edu/projects/getfile.asp?id=2830.

Tutumluer, E. Practices for Unbound Aggregate Pavement Layers. NCHRP Synthesis

No. 445, NCHRP Project 20-05, Topic 43-03, 2013.
https://trid.trb.org/view.aspx?id=1278501

Taylor, A. J. and D. H. Timm. Mechanistic Characterization of Resilient Moduli for

Unbound Pavement Layer Materials. NCAT Report-09-06, National Center for
Asphalt Technology, Auburn University, Alabama, 2009.

Woolstrum, G. (1990). Dynamic testing of Nebraska soils and aggregates. Transportation

Research Record, (1278). https://trid.trb.org/view.aspx?id=352766
 10

CHAPTER TWO: MANUSCRIPT ONE - LABORATORY CHARACTERIZATION OF

FINE-GRAINED SOILS FOR PAVEMENT ME DESIGN IMPLEMENTATION IN

IDAHO

Abstract
The subgrade layer often represents the weakest component of a pavement system

and can significantly affect pavement response and performance under loading. Adequate

characterization of subgrade properties is critical to the design and construction of long-

lasting, economical pavement systems. The mechanistic-empirical pavement design

protocol implemented through AASHTOWare® Pavement ME Design requires resilient

modulus of the subgrade soil as one of the primary input parameters. However, performing

repeated load triaxial tests to establish the resilient modulus properties of soils is a

cumbersome task, and is rarely performed by state transportation agencies on a regular

basis. In such cases, generic representative values are used by state transportation agencies

to specify the subgrade type during the pavement design process. Such simplifications and

gross estimations often fail to represent the actual soil types encountered in the field during

pavement construction. The Idaho Transportation Department (ITD) has recently invested

significant amounts of time and resources to facilitate state-wide implementation of

Mechanistic-Empirical (ME) pavement design practices. Under the umbrella of these

efforts, a research project was recently undertaken to characterize typical subgrade soil

types commonly encountered in the state of Idaho during pavement construction. Soil

samples were collected from different zones of Idaho, and extensive laboratory
 11

characterization of the soils was carried out including repeated load triaxial testing for

determining the resilient modulus values. One of the objectives was to establish a database

of typical soil properties that ITD engineers can use locally during pavement design and

performance prediction without having to rely on state-wide generic values. Soil properties

established through the laboratory testing effort were subsequently used as input in

Pavement ME Design to design typical flexible pavement sections, and to predict their

performances under traffic and environmental loading. The primary objective was to

quantify how laboratory characterization and establishing actual soil parameters without

using state-wide standard values could improve the overall pavement design practice. This

chapter, drafted in the form of a stand-alone manuscript, details findings from this project

and highlights the importance of such state-wide testing efforts to ensure effective

implementation of ME pavement design practices.

Introduction

The subgrade represents the weakest layer in a pavement structure, and is often

predominantly made of fine-grained soils. The response of these fine-grained soils under

traffic loading needs to be well-understood to facilitate the development and

implementation of effective pavement design methods. Subgrade soils, in particular weak

subgrades, can undergo significant deformations under repeated loading, thereby causing

significant damage to the overlying pavement structures. Plastic deformation

accumulations in the subgrade may contribute more than 40% of total rutting in a pavement

structure (Majidzadeh et al. 1978) Detailed characterization of the subgrade soil can help

in the design and construction of pavement foundations that perform well under traffic and

environmental loading. The Resilient Modulus (MR) is a critical property that governs
 12

unbound material behavior under loading, and is a critical input parameter during M-E

design of pavements. Resilient modulus represents the “elastic” stiffness of a material

under repeated loading once the initial plastic strain deformations have stabilized. This

manuscript focuses on resilient modulus characterization of eight different subgrade soil

types collected from across the state of Idaho, and evaluates how errors associated with the

quantification of subgrade resilient modulus can adversely affect pavement response and

performance prediction. Laboratory-established soil properties were used in

AASHTOWare Pavement ME design to predict the performances of pavement sections

constructed over these tested subgrade soil types. Pavement performance predicted through

such analyses were compared against those when state-wide standard values were used

instead as subgrade soil inputs.

Factors Affecting Resilient Response of Fine-Grained Soils

Accurate resilient modulus characterization is necessary to model the performance

and life span of a given pavement structure (Taylor 2008). Subgrade resilient modulus

primarily depends on stress states, soil physical properties and existing moisture conditions

(Rahim & George 2004; Oh et al. 2012; Nazzal et al. 2008). Nazzal et al. (2008) studied

resilient modulus for four groups of soils at four different moisture levels considering both

dry and wet side of optimum moisture contents and observed that resilient modulus

decreased by 50% to 70% with increasing moisture content. George (2004) found that

moisture content was the most prominent variables for predicting resilient modulus after

conducting sensitivity analyses on several models. Yau and Von Quintus (2002) performed

statistical analyses to identify soil properties that have significant effects on the resilient

modulus of soils. Their analysis showed that optimum moisture content was the most
 13

significant variable among other soil physical properties. Mishra et al. (2010) tested

aggregate sample at optimum moisture, dry of optimum moisture and wet of optimum to

study the effect of moisture contents. The overall effect of moisture on resilient modulus

was statistically insignificant at low fine contents. Recent investigation of subgrade soils

in Manitoba, Canada found high sensitivity in MR values because of moisture changes.

Those soils showed more than 5% permanent deformation when tested at wet side of

optimum moisture content (Soliman & Shalaby, 2010).

Pavement ME Design Implementation and Need for Correlation Development

The current flexible pavement design method adopted by the Idaho Transportation

Department (ITD) is an empirical procedure based on the subgrade R-value (Resistance

Value); this procedure grossly overestimates the required pavement layer thicknesses (El-

Badawy et al. 2011). Over the past few years, ITD has invested significant amounts of time

and resources towards state-wide implementation of Mechanistic-Empirical (ME)

pavement design practices. This has included projects focused on establishing relevant

input parameters for Hot Mix Asphalt (HMA) and Portland Cement Concrete (PCC)

materials. As the final component of this material characterization effort, a research study

was recently undertaken to characterize representative unbound materials (soils and

aggregates) commonly used in the state of Idaho for pavement applications. The primary

objective of this study was to generate a database of representative soil and aggregate

properties to be used as inputs during M-E pavement design. Another objective of the study

was to explore the adequacy of establishing correlations to accurately predict the resilient

modulus of subgrade soils from other easy-to-measure index/mechanical properties. This

is particularly important because repeated load triaxial tests (required to determine the
 14

resilient properties of unbound materials) are complex and time consuming in nature, and

it is often not feasible for state highway agencies to perform these tests in the laboratory

on a regular basis.

Objective of Study

The overall objective of this chapter is to establish resilient modulus characteristics

for different subgrade types commonly found across the state of Idaho, and quantify how

variations in resilient modulus properties affects pavement response and performance

prediction using AASHTOWare® Pavement ME Design. A supplementary objective was

to assess whether statistically significant correlations could be developed to predict the

resilient modulus of soils from other easy-to-measure index and mechanical properties.

The following tasks were performed towards accomplishing the overall project objectives:

• Perform repeated load triaxial tests following AASHTO T 307 specifications to

establish the resilient modulus characteristics of collected subgrade soil types

• Quantify the moisture sensitivity of subgrade soils by measuring the effect of

moisture content variations on resilient modulus, California Bearing Ratio (CBR),

and Unconfined Compressive Strength (UCS).

• Explore the feasibility of employing soil index properties and other easy-to-

measure mechanical properties to accurately predict soils resilient modulus.

• Study the effects of resilient modulus variations on pavement response and

performance predictions using AASHTOWare® Pavement ME Design.

• Study the effects of assumed MR of Idaho soils on critical pavement response

parameters, which were subsequently used in pre-establshed damage models to

predict flexible pavement life.

 15

Materials and Methods

This manuscript focuses on laboratory test results for eight different subgrade soil

materials (termed S1 through S8) that were tested in the laboratory. These materials

represent subgrade soils collected from different parts of Idaho. Note that although a total

of sixteen soil types were tested under the scope of the current research study, this

manuscript focuses on only eight soil types in an effort to reduce duplications among

similar soil types. The following sub-sections include results from the following laboratory

tests performed on the eight soil types: (1) grain size distribution and soil classification; (2)

compaction (moisture-density) characteristics; (3) California Bearing Ratio (CBR); (4)

Unconfined Compressive Strength (UCS); and (5) Resilient Modulus (MR) .

Grain Size Distribution

The first task in the laboratory test matrix involved establishing the particle size

distribution through sieving and hydrometer analysis. Table 2-1 lists percentage of gravel,

sand, silt and clay found in each soil sample from sieve and hydrometer analysis.
 16

Table 2-1 Percentage Gravel, Sand, Silt and Clay Content in Different Types of
Soils

Properties Soil Types

S1 S2 S3 S4 S5 S6 S7 S8

% gravel (76.2-4.75 mm) N/A 36.7 N/A 55.3 42.6 N/A 12 3

% sand (4.75-0.075 mm) 79.5 59.9 51.2 32.2 51.7 46.1 56.5 58.9

% fines (<0.075 mm) 20.5 3.4 48.8 12.5 5.7 53.9 31.5 38.1

% silt 17.8 3.1 43.7 10.7 5.1 47.2 27.0 34.9

% clay 2.7 0.3 5.1 1.8 0.6 6.7 4.5 3.2

As seen from the table, soil types S3 and S6 have approximately 50% fine particles

in their natural particle distributions and soil type S2 has the lowest amount of fine particles

among the eight soil types. Figure 2-1 shows the actual grain size distributions of soils.

Figure 2-1 Grain size distribution of different soil types

 17

Although the soil samples were collected from the subgrade layers for existing

pavement sections, some of the collected soil samples comprised significant amounts of

coarse particles; for example, soil types S2, S4 and S5. Note that the primary focus of this

component of the study was to characterize the fine-grained subgrade soils encountered

across the state of Idaho. Therefore, the research team, after discussions with ITD

engineers, decided to extract the fine fractions for the collected soil samples; all testing was

carried out on the fine fraction only.

Compaction Characteristics and Soil Classification

Moisture density characteristics for these subgrade soils were established in

accordance with the AASHTO T 99 test procedure, and the Maximum Dry Density (MDD)

and Optimum Moisture Content (OMC) values were established. The material fraction

passing through the 4.75 mm sieve was dried overnight in an oven conducting the moisture-

density tests. Commonly used soil index properties such as Atterberg’s limits (Liquid

Limit, Plastic Limit, and Plasticity Index) were also established. The soil classifications

were done according to the Unified Soil Classification System (USCS) and AASHTO

methods. Table 2-2 lists the liquid limit, plasticity index, MDD, and OMC values for each

of the eight soil types. Additionally, the USCS as well as AASHTO classifications for each

soil have also been listed in Table 2-2.

 18

Table 2-2 Soil Index Properties and Compaction Characteristics

Liquid Plasticity Index MDD OMC Soil Classification

Limit
Material PI ɤd,max ωopt
LL

% % kN/m3 % USCS AASHTO

S1 29 13 17.8 14 SC A-2-6

S2 N/A NP 20.1 8.5 SW A-1-b

S3 30 3 16.8 16.5 ML A-4

S4 32 3 17.3 15 GM A-1-a

S5 19 NP 21.1 8 SW A-1-a

S6 26 5 17.2 16.5 CL-ML A-4

S7 21 NP 19.3 11 SM A-2-4

S8 20 NP 18.5 12 SM A-4

Unsoaked California Bearing Ratio (CBR) Testing

CBR is an index property that can be linked to the shear strength of unbound

materials and was originally developed by the California Division of Highways to quantify

the shear strength of unbound materials (Croney & Croney 1991). It is a measure of the

resistance that a material presents in response to penetration of a standardized plunger.

Although the test is empirical in nature and does not necessarily represent any mechanistic

similarity with how the pavement layers are loaded, several states still actively use CBR in

their pavement design process. One common way is to use the CBR value to predict the

resilient modulus of an unbound material using empirically established equations. Two of

the most commonly used correlation equations are:

M R (psi) =
1500 × CBR for CBR ≤ 10 (Heukelom and Klomp, 1962)
= 2555 × CBR 0.64 (NCHRP 1-37A, 2004)
M R (psi)
 19

CBR tests were performed on unsoaked soil specimens following the AASHTO

T193 test protocol. Specimens were prepared at three different moisture contents to observe

the effects of moisture content on bearing capacity of soil. Specimens compacted at OMC

were used as the control specimens. Additionally, two specimens, one at 90% OMC and

the other at 110% OMC, were compacted and tested, simulating dry and wet conditions,

respectively. The test results have been tabulated in Table 2-3. As seen from the table, all

materials exhibited the highest CBR values under dry conditions; a reduction in CBR value

was observed when the moisture content increased to OMC, and subsequently to 110%

OMC.

Table 2-3 Unsoaked CBR Values for the Eight Soil Materials at Different
Moisture Contents

California Bearing Ratio (%)

Conditions S1 S2 S3 S4 S5 S6 S7 S8

0.9*OMC 27.5 61.4 23.7 36.1 18.1 16.9 60.4 36.8

OMC 6.7 55.8 5.2 12 12.3 3.4 38.1 14.3

1.10*OMC 3.5 10.5 1.6 2.0 3.3 2.1 9.2 3.2

Resilient Modulus Testing

Repeated Load Triaxial Tests (RLT) were conducted on the different soil types to

establish resilient modulus properties; the test procedure specified in AASHTO T307 was

followed. Haversine load pulses with 0.1s loading period and 0.9s rest period were applied

on top of cylindrical (100 mm diameter x 200 mm height) specimens. Six specimens were

prepared and tested for each material, constituting two replicates at the three different

moisture contents. For materials where significant differences in the test results were
 20

observed between the two replicates, a third specimen was tested, and results from all three

specimens were statistically analyzed before reporting an average value. The specimens

were prepared in three layers by applying 25 blows of the standard compactive effort

hammer. While preparing the specimens for resilient modulus testing, the target density

values for compaction were obtained from the moisture-density curves for the respective

materials. Vertical deformation of the specimen during repeated load triaxial testing was

measured using two externally mounted Linear Variable Differential Transformers

(LVDTs). The average value of resilient modulus from the last last five loading cycles

within each stress state was reported as the resilient modulus values for that particular stress

state. Table 2-4 lists resilient modulus values for the eight soil types under OMC

conditions. As already mentioned, two replicates were tested for each soil type. Therefore,

the values listed in Table 2-4 represent average of the values obtained from the two

replicates.
 21

Table 2-4 Resilient Modulus of Soils at Different Stress States Tested at OMC

Seq. Confining Axial Resilient Modulus of different Materials

No. Pressure Stress S1 S2 S3 S4 S5 S6 S7 S8

kPa kPa MPa MPa MPa MPa MPa MPa MPa MPa

1 41.4 12.4 79.6 71.7 39.5 103.5 55.6 45.5 31.9 24.5

2 41.4 24.8 71.2 72.7 35.9 88.5 62.7 41.6 39.2 29.7

3 41.4 37.3 61.4 70.6 32.7 78.4 65.9 35.3 42.1 32.6

4 41.4 49.7 51.7 71.2 31.5 72.3 68.7 29.2 46.4 36.1

5 41.4 62.0 46.5 75.1 34.1 71.0 74.4 27.6 52.7 40.0

6 27.6 12.4 61.7 46.6 21.9 65.1 38.9 29.8 24.2 19.1

7 27.6 24.8 50.9 49.2 20.8 53.2 49.3 24.6 30.3 23.6

8 27.6 37.3 45.2 56.8 23.8 52.7 60.1 23.1 38.4 29.6

9 27.6 49.7 43.2 64.9 27.8 55.4 67.7 23.5 46.6 35.5

10 27.6 62.0 42.8 71.3 32.0 58.2 73.5 25.1 53.5 40.1

11 13.8 12.4 53.9 35.6 16.7 38.3 35.6 25.1 22.1 17.3

12 13.8 27.6 43.8 42.4 18.5 36.3 48.1 20.6 28.9 22.6

13 13.8 41.4 39.2 52.1 21.2 40.3 60.1 19.9 37.2 28.9

14 13.8 55.2 38.1 61.1 26.7 45.6 68.1 21.2 45.9 35.3

15 13.8 68.9 38.5 67.9 28.1 50.3 73.8 23.4 53.3 40.4

Unconfined Compressive Strength

Unconfined compressive strength tests were performed on the eight soil types in an

effort to identify whether statistically significant correlations existed between resilient

modulus and UCS values. The same specimen used for resilient modulus testing was later

tested for UCS. A total of 48 specimens (8 soil types x 3 moisture contents x 2 replicates)

were tested for UCS under controlled strain conditions at a rate of 1 mm/minute (as
 22

specified in AASHTO T208). The peak axial stress attained for each specimen was

recorded, and the average value from the two replicates was reported for a given soil type

and moisture condition combination, the results have been tabulated in Table 2-5. As seen

from the table, all soils exhibited significant moisture sensitivity with UCS values rapidly

decreasing with increasing moisture content.

Table 2-5 Peak Compressive Strength of Soils at Different Moisture Contents

Moisture Unconfined Compressive Strength (kPa)

Conditions
S1 S2 S3 S4 S5 S6 S8 S9

0.9*OMC 248.9 113.8 202.7 178.6 134.4 220.6 172.4 165.5

OMC 162.7 104.8 155.1 114.5 96.5 166.2 117.2 144.8

1.10*OMC 75.8 82.7 120.7 113.8 48.3 110.3 103.4 51.7

As already mentioned, one of the objectives behind running the UCS tests was to

try and establish statistically significant correlations between UCS and MR test results. Lee

et al. (1997), Drumm et al. (1990), and Hossain et al. (2011) reported good correlations

between UCS and MR, and successfully developed correlation equations to predict MR from

UCS test results.

Analysis of Test Results

Resilient Modulus Model Parameters

MEPDG input level-1 requires constitutive model parameters to calculate material

responses. The constitutive model used in ME pavement design is widely known as

universal model or MEPDG model (AASHTO, 2008). This model can incorporate the

effect of both deviatoric and volumetric stress. The required values from repeated load
 23

triaxial testing were used in Equation (1) and multiple linear regression analysis was

performed to determine model parameters.

k2 k3
 θ  τ 
M R = k1 × pa ×   ×  oct + 1 (1)
 pa   pa 

where
θ is known as bulk stress or the first stress invariant (θ = σ 1 + σ 2 + σ 3 )
σ 1 is the major principal stress
σ 2 is the intermediate principal stress
σ 3 is the minor principal stress
For a cylindrical triaxial test set-up
σ2 = σ3
σ=
1 σ3 +σd
pa is atmospheric pressure, and equals 101.3 kPa
The pa term is used for normalization in the equation
τ oct represents the octahedral shear stress
1
τ oct= (σ 1 − σ 2 ) + (σ 2 − σ 3 ) + (σ 3 − σ 1 )
2 2 2

3
For a cylindrical triaxial test specimen,
2 2
τ oct = (σ 1 − σ 3 ) = × σ d
3 3

In Equation (1), k1 values are always positive, and can be thought to be

representative of the Young’s modulus of the material. Positive values for the k2 parameter

would mean that increase in the bulk stress leads to stiffening of the material. Similarly,

negative values for the k3 parameter would imply that increase in the shear stress leads to

softening of the material. Table 2-6 lists parameters for the eight different soil types after

the universal model was fitted to the resilient modulus test results.
 24

Table 2-6 MEPDG Model Parameters for the Eight Soil Materials
Corresponding to Three Different Moisture Contents

Material 0.9*OMC OMC 1.10*OMC

ID

2 2 2
k1 k2 k3 R k1 k2 k3 R k1 k2 k3 R

S1 1346.5 0.34 -1.93 0.95 767.2 0.52 -3.08 0.90 − − − −

S2 1182.4 0.69 -1.74 0.95 486.9 0.55 0.53 0.84 283.6 0.26 2.17 0.89

S3 768.4 0.67 -2.61 0.81 260.6 0.76 -0.72 0.77 − − − −

S4 1001.4 0.84 -2.72 0.91 744.8 0.99 -2.47 0.89 263.5 0.81 0.07 0.78

S5 410.2 0.23 2.35 0.92 397.6 0.26 1.94 0.85 − − − −

S6 555.9 0.57 -2.66 0.86 377.2 0.68 -2.68 0.76 − − − −

S7 327.4 0.17 1.58 0.94 223.6 0.25 2.78 0.94 190.1 0.18 3.51 0.91

S8 273.9 0.24 1.25 0.89 175.2 0.24 2.70 0.94 − − − −

The first thing to be noticed from the above table involves the missing model

parameter values for several soils when tested under the wet conditions (110% OMC). This

was primarily because the specimens were too weak at the high moisture content, and could

not sustain the repeated loading during resilient modulus testing. Next, it should be noted

that the k1 and k2 parameter values were positive for all soil materials under all three

moisture contents. The k1 parameter is always expected to have a positive value because a

negative value would indicate a negative modulus for the material, which has no physical

meaning. Positive values for the k2 parameter indicate that all soil materials at all moisture

content attained higher modulus values as the bulk stress value was increased. The k3 model

parameter on the other hand, was positive for some soils and negative for some. Although

fine-grained soils are expected to exhibit softening behavior under increasing shear stress

levels, the variation in the sign of the parameter can be attributed to error minimization
 25

techniques during multiple linear regression. Note that the model fitting was carried out

with the primary objective of minimizing the error between observed and predicted

resilient modulus values, and no physical meaning to the model parameters were assigned

during the regression process. No mechanistic justification can be provided for the positive

k3 values obtained for some of the soils.

Effect of Moisture Content and Deviatoric Stress on Soil Resilient Modulus

The effect of changing moisture content on MR was thoroughly investigated in this

study. To aid the discussion of these test results, the eight soils have been classified into

two categories: (1) plastic soils; and (2) nonplastic soils. Figure 2-2 (a)-(d) show the

variation in resilient modulus for soils S1, S3, S4, and S6, respectively, with applied

deviatoric stress levels. As can be seen from the Atterberg’s limit tests presented earlier in

this chapter (Table 2-2), these soils exhibited non-zero Plasticity Index (PI) values, thereby

being classified as “Plastic”. On the other hand, soils S2, S5, S7, and S8 had zero PI values,

thus being classified as “nonplastic”; results for these soils have been presented in Figure

2-3 (a)-(d). Note that the different lines in the plots correspond to different confining

pressures, and also to different moisture conditions. Remember that AASSHTO T 307

protocol requires each specimen to be subjected to five different deviator stress levels at

three different confining pressures, thus resulting in 15 different stress states.

 26

(a) (b)

(c) (d)

Figure 2-2 Variation of MR with different moisture content for plastic soils

Careful inspection of the above figure clearly highlights that that moisture content

had a significant impact on the MR values for the soils with non-zero PI values; significant

reduction in MR was observed as the moisture content increased. This trend was consistent

across all confining pressure levels. It can also be seen from the plots that as the deviator

stress magnitude increased, the MR values mostly decreased, implying stress-softening

behavior. This trend was reversed in certain cases, for example, the wet condition for Soil
 27

S4 (Figure 2-2c). Increasing MR with increasing deviator stress implies stress-hardening

behavior, and should lead to a positive k3 parameter for this particular soils. This is

confirmed from Table 2-6 (soil S4 has a positive k3 parameter under wet conditions). It

should also be noted that three of the four plastic soils failed during resilient modulus

testing when the specimen was compacted at 110% OMC. This finding agrees with the

trend observed by Mishra (2012), where he reported that unbound aggregate specimens

with plastic fines exhibited significantly higher moisture sensitivity compared to those with

nonplastic fines.

Figures 2-3a through 2-3d present similar results for the remaining four soil types,

that were grouped under the ‘nonplastic’ category. As seen from these plots, the soils under

the nonplastic category did not exhibit as significant of a drop in resilient modulus with

increasing moisture content except for soil S2. More importantly, three (S5, S7, and S8) of

the four soil types exhibited increasing trends in MR values with increasing deviator stress

magnitude. Even for soil S2, this trend was observed for the wet conditions. This clearly

shows that in most cases, the soils with nonplastic fines exhibited stiffening response as

the deviator stress magnitude was increased. This is demonstrated by positive values for

the k3 parameter for all these cases except for the case when soil S2 was tested under dry

conditions.
 28

(a) (b)

(c) (d)

Figure 2-3 Variation of MR with different moisture content for non-plastic soils

From the above results, it can clearly be seen that although moisture content

adversely affects the resilient modulus values for all soil types, that effect is much more

pronounced in case of soils with plastic fines.

 29

Predicting Soil Resilient Modulus from Index Properties

As already mentioned, one of the objectives of this research effort was to evaluate

whether or not resilient modulus values for subgrade soils could be predicted with

reasonable accuracy from easy-to-determine index and mechanical properties. This would

help state and local highway agencies acquire relevant subgrade soil input parameters for

mechanistic-empirical pavement design without having to run cumbersome and resource-

intensive repeated load triaxial tests. The current research study utilized soil gradation

information, index properties, compaction characteristics, as well as CBR and UCS values

to try and develop correlation equations to predict the resilient modulus value. However,

note that resilient modulus is a stress-dependent property and cannot be specified using a

constant value for a particular material. Nevertheless, the current version of the

AASHTOWare Pavement ME Design software requires a single value of resilient

modulus as input. To help the reader overcome this dilemma, some discussion on how to

arrive at a single resilient modulus value for a soil is warranted. This discussion is presented

in the following subsection.

Specifying One Resilient Modulus Value for an Unbound Material

Resilient modulus is a stress-dependent parameter and can take up different values

depending on the stress state that a particular unbound material (soil or aggregate) is

subjected to. A material that is very well-confined, will usually be stiffer compared to a

material that is not well-confined. This is particularly relevant in a pavement structure

because the confinement levels as well as vehicle-imposed stress levels change spatially

within an unbound layer. An element closer to the pavement surface is subjected to much

higher stress levels (from vehicular loading) compared to an element further down in the
 30

pavement system. This is also true within a particular layer. The vehicle-imposed stresses

are much higher closer to the point of loading and dissipate gradually as we move away

vertically or laterally. Considering that resilient modulus is highly influenced by the stress

conditions, it is expected that the resilient modulus of an unbound layer will exhibit

significant spatial variation.

Traditional pavement analysis approaches use the linear elastic theory where each

layer, including unbound soil and aggregate layers, is represented by a single modulus

value. Although this can be thought of as an over-simplification of actual material behavior,

considering spatially variable modulus values within a layer requires the use of

computationally expensive approaches such as the finite element method. Although several

advanced pavement analysis codes such as ILLI-PAVE (Thompson & Elliott, 1985) and

GT-PAVE (Tutumluer, 1995) have been developed with the capability to consider stress-

dependent (and hence spatially variable) modulus values for soil and aggregate layers,

using such analysis approaches during ME design would render the design and analysis

process excessively time consuming. Therefore, the ME analysis and performance

prediction approach implemented in AASHTOWare Pavement ME Design makes use of

JULEA, a layered-elastic analysis program developed by Uzan (1976). This in turn requires

the user to specify a single modulus value to be used during the design process. Note that

this departure from actual soil and aggregate behavior has been deemed to be acceptable

because the stress-induced variations in soil and aggregate modulus is not likely to affect

the pavement response and performance as significantly as other factors such as variations

in HMA layer thickness/modulus and/or variations in climatic conditions.

 31

In an effort to propose a single modulus value that can be used to represent the

entire soil/aggregate layer, NCHRP project 1-28 (Barksdale et al. 1997) proposed the

concept of ‘Summary Resilient Modulus’. This approach recommended using the resilient

modulus value obtained from laboratory testing at 14 kPa confining pressure and 41 kPa

cyclic stress as the ‘Summary Resilient Modulus (SRM)’ for fine-grained subgrade soils.

This stress state is assumed to be representative of the typical stress field experienced by a

subgrade element in a pavement section under vehicular loading. Therefore, the current

study used the concept of summary resilient modulus to define one single modulus value

for each of the soils tested. Note that the stress states used for defining the SRM for fine-

grained soils correspond to stress sequence number 13 during AASHTO T 307 testing.

Correlation Development to Predict Summary Resilient Modulus Values

Once the decision was made to use the SRM as the representative modulus value

for each tested soil, the first task involved analyzing the trends in SRM values under

different test conditions. A Box-and-Whisker plot was developed for all the SRM values

corresponding to different moisture conditions (see Figure 2-4). Note that the horizontal

line within the shaded boxes in represents the median value for that particular distribution.

The boundaries of the box correspond to the 25th and 75th percentiles, respectively. The

straight line beyond the box extend to show the scatter of data for that particular moisture

condition.
 32

Figure 2-4 Distribution of summary modulus data tested at each moisture

content

As seen from the plot, the median SRM value for specimens tested under dry

conditions was the highest, whereas the median value corresponding to wet conditions was

lowest. Note that the median value for the wet conditions lies at ZERO because several of

the specimens (more than 50%) failed during resilient modulus testing, and their SRM

value was taken to be zero for this analysis. It is important to note that although the SRM

value reduces as the moisture condition is increased from dry condition to OMC, the most

significant drop in SRM occurs when the moisture content is increased beyond OMC to

wet conditions. This wide variation in SRM value with changing moisture content clearly

implies that any equation developed to predict the SRM value must take the moisture

content of the specimen into consideration. Extensive statistical analysis of the test data

was carried out to explore whether it was possible to develop generalized correlation

equations to predict the SRM values at different moisture contents. However, it was

observed that no such generalized equation resulted in statistically significant correlations.

 33

Therefore, the discussion in this chapter focuses on the development of correlation

equations to predict the SRM when the moisture content equals OMC.

Multiple regression analysis was carried out using the R statistical package (Team,

2013) to develop correlations between SRM and different soil index and mechanical

properties. The model equations for predicting SRM are shown in Equation (2) and

Equation (3) where independent variables are percent finer than #200 sieve (P#200),

compaction moisture content (W), unconfined compressive strength (qu), and percentage

clay present (% clay).

MR = 69.41 − 0.51× P # 200 − 1.44 × W (2)

MR = 74.28 − 0.18 × qu − 3.99 × %Clay (3)

Close inspection of the two equations presented above leads to several interesting

observatios. For example, the first equation (correlating Mr with ‘P200’ and ‘W’) has

negative coefficients associated with the two independent variables. This means, the SRM

value decreases as both P200 and W values are increased. This is consistent with commonly

observed soil behavior, as both P200 and W have been shown to have negative correlations

with resilient modulus. The second equation, on the other hand, proposes relationships that

do not agree with engineering observation. For example, both the independent variables

(qu and %clay) in the second equation are associated with negative coefficients. Although

reduction in SRM with increasing %clay seems logical, the negative coefficient for qu has

no physical significance. There is no scientific reason for SRM to have a negative

correlation with qu. Therefore, it is quite clear that the above-mentioned equation is merely

an outcome of an error minimization process, and does not have any significant
 34

mechanistic standing. Therefore, the implementation of such an equation into practice

cannot be justified.

Besides attempting to develop correlations equations for predicting the SRM

values, the current study also tried to develop correlation equations to predict the MEPDG

model (also known as the universal model) parameters (k1, k2, and k3) using the index and

other mechanical properties. If the model parameters can be predicted with reasonable

accuracy, this would help in calculation of representative resilient modulus values if typical

stress states experienced by a soil element in a pavement system can be estimated. The

equations for predicting MEPDG model parameters are shown in Equation (4), Equation

(5) and Equation (6) below.

K1 =
−93.1 − 15.5 × P # 200 + 73.98 × W (4)

K2 =
−0.47 − 0.01× P # 200 + 0.10 × W (5)

K 3 = 8.95 + 0.10 × P # 200 − 0.93 × W (6)

Statistical parameters indicating the quality of these regression equations were obtained

from R, and have been listed in in Error! Reference source not found.. The parameters

that were considered to measure the quality of the fit were: residual standard error (RSE),

F statistics, model significance in term of P-value and precision of developed model in

term of Adjusted R-square.

 35

Table 2-7 Multiple Regression Analysis Output for the Five Models

Models RSE DF F-value P-value Adj. R-squared

Equation (2) 3.356 5 59.04 0.00033 0.94

Equation (3) 5.855 5 17.71 0.0054 0.83

Equation (4) 96.86 5 16.48 0.0063 0.82

Equation (5) 0.1464 5 9.62 0.0193 0.71

Equation (6) 1.339 5 9.26 0.0208 0.70

Once again, it is important to closely inspect the equations resulting from regression

analysis to assess whether or not they carry any physical meaning. Taking a look at the first

equation (to predict k1), it can be seen that the intercept has a negative value, and the

coefficients associated with ‘P#200’ and ‘W’ are negative and positive, respectively. This

is questionable at several levels. Firstly, based on this equation, if a soil has zero percent

of particles passing the #200 sieve (finer than 0.075 mm), and is compacted under

completely dry conditions (moisture content close to zero), then the k1 value will be

negative. As the k1 parameter is supposed to correspond to the Young’s modulus of the

material, a negative k1 parameter would mean a negative Young’s modulus, which has no

meaning whatsoever. Additionally, the equation shows that k1 value would increase with

increasing moisture content; this is also erroneous as increasing the moisture content

(particularly on the wet side of OMC) leads to significant weakening of soils. Similar

limitations can be easily identified for the equations to predict k2 and k3. Therefore, it can

be clearly seen that the equations to predict the MEPDG model parameters, though lead to

reasonably good statistical significance, have no physical meaning associated with them.
 36

Therefore, using these equations to predict the MR for soils will lead to erroneous pavement

designs.

Effect of Soil Resilient Modulus on Predicted Pavement Performance Using

AASHTOWare Pavement ME Design

The ME pavement design procedure presents the user with the option to perform

the analysis at three different levels based on the amount of information available. Level 1

corresponds to the case when all design inputs have been established experimentally. Level

3 on the other hand corresponds to the case when very little information is available

regarding the design input data, and nation-wide assumed/default values are used during

the design process. Level 2 presents an intermediate scenario as far as the availability of

design data is concerned. As far as resilient modulus value of soil is concerned, level 1

corresponds to the case where the user has data from laboratory resilient modulus tests

conducted for the particular subgrade soil type under consideration (currently,

AASHTOWare Pavement ME Design does not have the ability to incorporate level 1

inputs for unbound materials). Level 3 on the other extreme corresponds the case where no

laboratory test data is available for the soil, and the resilient modulus needs to be assumed

based on generic information such as the USCS or AASHTO soil classification. Level 2

lies between these two extremes and represents a case where although MR test results are

not available, data from certain other tests such as CBR, UCS, R-value, etc. are available,

and can be used to predict the resilient modulus values through commonly accepted

correlation equations.

The previous section clearly proved that no correlation of statistical significance

and following commonly observed material behavior trends could be developed to predict
 37

soil resilient modulus from index and other commonly measured mechanical properties.

The next task involved evaluating the effect of soil resilient modulus on pavement

performance prediction using Pavement ME Design. The objective was to compare

predicted pavement performance when laboratory-measured soil properties are used as

design inputs in place of commonly made assumptions. If the difference in predicted

pavement performance is found to be significant, then state highway agencies would be

recommended to somehow compile a database with actual resilient modulus test data for

typical soil types rather than using previously established correlation equations.

This task was completed by randomly selecting a state highway in Idaho for case

study. Traffic and other relevant inputs required by AASHTOWare Pavement ME Design

(version 2.3.1) were obtained from the User’s Guide developed for ITD by Mallela et al.

(2014). A conventional flexible pavement section was designed comprising HMA, base,

subbase, and subgrade layers. Two different designs were performed using two differen

types of subgrade inputs. The first design assumed no first-hand information about the

subgrade properties and used the default values recommended by Mallela et al. (2014). The

second design used the test data from the current laboratory testing effort. Results from the

two cases have been discussed in the following sections. Note that during the discussions,

the first design has been referred to as “Recommended Subgrade” indicating that this

design alternative uses subgrade properties as recommended by the User’s guide developed

for ITD by Mallela et al. (2014); the second design has been termed as the “Tested

Subgrade”. The traffic data was obtained from ITD’s Weigh-in-Motion (WIM) records.

Note that material properties for the HMA, Base, and subbase layers, as well as traffic
 38

related inputs were kept unchanged between the two cases; the only difference was in the

subgrade properties.

At first, suitable layer thicknesses were established by using all inputs as

recommended by the User’s Guide (Mallela et al. 2014). While converging on the layer

thickness values, it was ensured that the resulting pavement section did not fail in any

category during its design life (for the specified design reliability level). Minimum required

layer thicknesses obtained from this approach have been marked in Figure 2-5.

Figure 2-5 Minimum required layer thicknesses while using all ITD-
recommended values as material inputs.

Once the minimum required layer thicknesses were established, the analysis and

performance prediction process was repeated by replacing the ITD-recommended material

property values for the subgrade with those obtained from the laboratory testing carried out

in this study. Note that all other aspect of the pavement structure (layer thicknesses as well

as loading conditions, etc.) were unchanged from the first case. Performance predictions

for the two design cases have been compared in Table 2-8.
 39

Table 2-8 Comparison of Pavement Performances Obtained from Pavement

ME.

Target
Recommended Subgrade Tested Subgrade
values at
90%
reliability Predicted % Reliability Predicted % Reliability

Terminal IRI (mm/km) 2762 1850 99.99 2020 99.89

Total Permanent
12.7 12.7 90.41 18.5 12.68
Deformation (mm)

AC bottom up fatigue
15 1.45 100 1.45 100
cracking (% lane area)

AC thermal cracking
284 51.2 100 51.2 100
(m/km)

AC top-down fatigue
947 57.2 100 55.5 100
cracking (m/km)

AC-permanent deformation
12.7 2.54 100 2.54 100
(mm)

As seen from the above table, several of the performance criteria such as: Terminal

IRI, AC bottom up and top down fatigue cracking, AC thermal cracking and AC permanent

deformation were within target reliability levels for pavement structures built using both

ITD-recommended as well as laboratory-established subgrade soil properties. However, it

can be seen that the total predicted permanent deformation value changed significantly

(from 12.7 mm to 18.5 mm) when the assumed soil properties were replaced with those

established through laboratory testing. The limiting criterion for total permanent

deformation was met in the first design case, whereas it failed for the second design case.

Note that the Guide-recommended subgrade soil modulus values was 103 MPa, whereas a

value of 40 MPa was used as the Summary Resilient Modulus (SRM) from the laboratory

testing results. It is quite evident that using a lower modulus value for the subgrade soil

resulted in a significantly higher predicted value for the total permanent deformation.
 40

Figure 2-6 shows the accumulation in total permanent deformation with pavement age for

the two design cases. The pavement structure using ITD-recommended values as inputs

(design case 1) did not exceed the limiting value (12.7 mm) for total permanent

deformation accumulation during the pavement’s lifetime. On the other hand, the pavement

structure with laboratory-established soil input properties (design case 2) exceeded the

limiting value for total permanent deformation accumulation within three years of

construction (for 90% design reliability). Even for the case with 50% design reliability, the

limiting value for total permanent deformation accumulation is exceeded within six years

of construction.

(a) (b)

Figure 2-6 Predicted rutting for (a) recommended subgrade (b) tested subgrade

Effect of Soil Resilient Modulus on Critical Pavement Response Parameters

The pavement performance prediction approach inherent to M-E design relies on

complex transfer functions that use relevant pavement responses, such as stresses, strains,

deflections, etc., along with several other variables to predict pavement performance. Note

that the nature of these transfer functions is often not evident to the user. Therefore, this
 41

study directly compared critical pavement response parameters for the different pavement

sections to discuss how the performances are likely to be affected. Critical pavement

response parameters for different pavement sections can be directly compared to

categorically isolate the effects of certain factors of interest on pavement behavior. The

critical pavement response parameters considered were: (1) Horizontal strain at bottom of

HMA layer (HStrn-HMA); (2) deviatoric stress on top of the subgrade (DStrss-Sub); and

(3) vertical strain on top of the subgrade (VStrn-Sub). Simple damage models such as the

ones proposed by the Asphalt Institute would help translate the critical pavement response

parameters into performance (in terms of number cycles to failure).

Note that AASHTOWare Pavement ME Design is based on JULEA (Jacob Uzan

Layered Elastic Analysis) as the background analysis engine, and is therefore based on the

multi-layer linear elastic theory. In its current version, the software does not account for

stress-dependent unbound material behavior (Ceylan & Gopalakrishnan 2011).

To quantify how resilient modulus value affects critical pavement response

parameter calculations, the current study used the commonly used layered elastic analysis

program: KENLAYER (Huang, 1993). The pavement section shown in Figure 2-5 was

used for this analysis, and the input parameters were kept similar to those used during the

Pavement ME Design runs. The analysis case that used ITD-recommended subgrade soil

properties has been referred to here as the ‘ITD recommended pavement section’. The

alternate analysis case used laboratory-generated subgrade soil properties and is referred

to as the ‘Tested subgrade’ case. A standard wheel load with 552 kPa contact pressure and

150 mm radius was applied from top of the HMA layer. The critical pavement response
 42

parameter values were extracted at different points within the pavement structure and have

been plotted in Figure 2-7.

(a) (b)

(c)

Figure 2-7 Variation in predicted pavement performances; a) vertical

displacement b) vertical stress and c) horizontal strain with depth starting from top
of pavement.

Remember that the ITD-recommended subgrade case was assigned a higher

resilient modulus values compared to the tested case. Accordingly, all layers in ITD-

recommended pavement structure had lower vertical deformation magnitudes compared to

the tested case. The vertical stress magnitudes were found to be relatively similar for the

two cases. The horizontal strain at the bottom of the asphalt layer was found to be quite
 43

similar for the two cases. Note that Figure 2-8c shows tensile strain in the unbound layers,

which is not possible in the real world. This is nothing but an outcome of layered-elastic

analysis approach, which treats every layer as a continuum with equal modulus in

compression as well as tension. From the above figure, it can clearly be seen that changing

the subgrade modulus is likely to primarily affect the rut accumulation in the pavement (a

function of the vertical strain/displacement. Other distresses such as fatigue cracking are

not likely to be affected significantly by changes in subgrade modulus.

Use of Simple Damage Models to Compare Pavement Life

Once the critical pavement response parameters were calculated for the two

alternate designs, the next step involved using simple damage models proposed by the

asphalt institute to compare the number of cycles to failure for these pavement sections

under fatigue and rutting criteria. The horizontal tensile strain at the bottom of HMA layer

was used in fatigue cracking model shown in Equation (7) to predict number of load

repetitions. Similarly, vertical compressive strain at the top of subgrade were used in

Equation (8) to predict number of load repetitions. Both of the equations were proposed by

Asphalt Institute for damages analysis.

−0.854
N f 0.0796(∈t )−3.291 E *
=
(7)

N d = 1.365 ×10−9 (∈c )−4.477 (8)

Table 2-9 lists the critical pavement response parameters and the corresponding load

repetitions to failure calculated using the damage models for rutting and fatigue cracking.
 44

Table 2-9 Sensitivity of Subgrade Resilient Modulus in Damage Predictions

Subgrade Subgrade soil Tensile strain Vertical strain Load Load

Modulus repetitions repetitions
Types

MPa ∈t ∈c N f
Nd

ITD 103 -1.25×10-4 1.757×10-4 5.35×106 88.56×106

Recommended

Tested 40 -1.295×10-4 2.794×10-4 4.765×106 11.10×106

Subgrade

Results from the rutting model calculations show that ITD-assumed subgrade

modulus allows 8 times greater number of load repetitions compared to the tested subgrade

case. Similarly, calculations from the fatigue cracking model estimated 12.3% higher

number of load repetitions before failure for the pavement with assumed soil modulus

properties. These observations clearly highlight the importance of laboratory testing to

establish resilient modulus properties of subgrade soils rather than using nationally

assumed standard values.

Effect of Moisture-Induced Subgrade Resilient Modulus Variation on Pavement

Performance Prediction

As already mentioned, the subgrade soils in the current study were tested under three

different moisture conditions (90% of OMC, OMC, and 110% of OMC) to evaluate their

moisture sensitivity. The next task in the pavement analysis and performance prediction

using AASHTOWare Pavement ME Design involved determining minimum required

pavement layer thicknesses to ensure adequate pavement performance when the subgrade

moisture content varied between these extremes. These variations in subgrade moisture can

be due to seasonal effects, or due to inadequate drainage conditions that cause unexpected
 45

ingress of water into the pavement structure. As is common practice, the User’s Guide

developed for ITD by Mallela et al. (2014) recommends using subgrade properties under

OMC conditions. However, it is quite possible for the subgrade moisture condition to

deviate significantly from OMC conditions, which would lead to different required

pavement layer thicknesses. The following paragraphs present results from this analysis.

Subgrade soil at OMC conditions was treated as the control case in this analysis. Two

other design cases, one with the subgrade moisture content set to 90% of OMC, and the

other at 110% of OMC were considered, and the minimum required pavement layer

thicknesses were compared. Note that properties for the HMA, base, and subbase layers,

as well as the traffic inputs were kept unchanged for all three cases. Required pavement

layer thicknesses for the three different subgrade moisture conditions have been listed in

Table 2-10.

Table 2-10 Required Layer Thickness of Pavement Built in Tested Subgrade

Layer type Layer Thickness (mm)

Control Dry Wet

HMA 190 173 216

Base 254 203 254

Subbase 305 305 305

From the above table it can clearly be seen that when the subgrade moisture condition

was set to “wet”, the HMA layer thickness needed to be increased (from 190 mm to 216

mm) to satisfy the preset performance criteria. On the other hand, when the subgrade

moisture condition was set to “dry” reductions in HMA layer (from 190 mm to 173) as

well as base layer (from 254 mm to 203 mm) thicknesses could be accommodated while
 46

still meeting all performance criteria. This clearly highlights the importance of preventing

moisture ingress into the subgrade layer if adequate pavement performance is to be

maintained.

The next step in this analysis involved studying the case where the pavement section

has been designed assuming OMC conditions for the subgrade, but a significant fluctuation

in the moisture content occurs during the design life of the pavement. To study this

scenario, all layer thicknesses were kept constant at the values for the control case

(subgrade moisture at OMC); only the subgrade modulus value was altered to simulate

moisture content variations. Figure 2-8 shows the variation of IRI (Figure 2-7a) and Rut

Depth (Figure 2-7b) with age for a flexible pavement structure built on subgrade soils

compacted at three different moisture conditions.

(a) (b)

Figure 2-8 Effect of subgrade moisture on a) IRI and b) rut depth of flexible
pavement

Higher moisture contents result in lower modulus values which ultimately lead to

higher IRI values. Among the three moisture conditions being analyzed, the pavement

section with dry subgrade corresponded to the lowest predicted IRI value. Although the

IRI values predicted for pavement sections with OMC and wet subgrade conditions were
 47

higher than that for the one with dry subgrade, it should be noted that all three pavement

sections passed the IRI performance criteria. As seen from Figure 2-7b, the pavement

structure built over wet subgrade exceeded the pre-established threshold value for rutting

within seven years after construction; on the other hand, the pavement section with dry

subgrade did not reach the threshold life till the end of its design life. Excess moisture in

the subgrade soil makes it weaker which results in higher permanent deformation

accumulations under traffic loading. From the analysis results it was clear that rut depth

predictions changed significantly with changing subgrade moisture conditions, whereas

predictions for several other performance indicators such as bottom up cracking and

thermal cracking remained unchanged. Note that the effect of moisture content on IRI,

although noticeable, was not as significant as that for rut accumulation.

Effect of Soil Moisture on Critical Pavement Response Parameters – Results from

Finite Element Analysis

The above section discussed the effects of changing soil moisture conditions on

pavement performance prediction using AASHTOWare Pavement ME Design. However,

as already mentioned, rather than comparing the predicted performances, sometimes a

better understanding of pavement behavior can be obtained by comparing the critical

pavement response parameters. This was accomplished in the current study by analyzing

the pavement sections using ILLI-PAVE, a finite element-based pavement analysis

program that is capable of considering stress-dependent unbound material behavior.

Developed at the University of Illinois by Thompson and Elliott (1985), ILLI-PAVE

considers stress hardening behavior of aggregates and stress softening behavior of fine-

grained soils. The ILLI-PAVE analyses carried out in this study would help determine
 48

whether ignoring the stress-dependent behavior of unbound materials is likely to affect the

performance predictions significantly or not.

ILLI-PAVE 2005 (version 5.0) finite element tool uses deviator stress dependent

bilinear approximation for subgrade soils, and and the K-θ model for unbound aggregates

to account for stress-dependent material behavior (Tutumluer 1995). The laboratory-

generated resilient modulus test results were fitted with adequate models, and the model

parameters were used as inputs for ILLI-PAVE analysis. Three different sets of resilient

modulus modulus parameters were considered for the subgrade soil corresponding to the

three different moisture conditions. The pavement section with subgrade corresponding to

OMC conditions was treatd as the ‘control section’. Values for the critical pavement

response parameters were extracted from the ILLI-PAVE output, and have been listed in

Table 2-11.

Table 2-11 Predicted Pavement Responses Found from ILLIPAVE Stress

Dependency Tool

Critical Responses Control Section Dry Subgrade Wet subgrade

Values Depth mm Values Depth Values Depth

mm mm

HStrn-HMA 137.6x10-6 216 131.9x10-6 216 151.6x10-6 216

DStrss-Sub (kPa) 13.1 483 13.8 483 9.6 483

VStrn-Sub 341.5x10-6 483 259.2x10-6 483 551.9x10-6 483

*HStrn-HMA: Horizontal Strain at the bottom of the asphalt layer

**DStrss-Sub: Deviator Stress at the top of the subgrade
***VStrn-Sub: Vertical Strain on top of the subgrade

As seen from the table, the subgrade soil under dry conditions leads to lower tensile

strain values at the bottom of HMA layer and lower vertical strain on top of soil surface. It
 49

can clearly be seen that the change in critical pavement response parameter magnitudes is

not as drastic when the moisture conditions are changed from OMC (control) to dry

conditions compared to the case when moisture content changes from OMC to wet

condition. The wet subgrade condition results in the highest (elastic) vertical strain on top

of the subgrade, which in turn can be linked to the highest rut susceptibility.

Summary and Conclusions

The manuscript focused on laboratory characterization of typical subgrade soil

types encountered in the state of Idaho to facilitate state-wide implementation of M-E

pavement design practices. Eight different soil materials were collected from different parts

of Idaho and were tested in the laboratory. Preliminary laboratory tests such as grain size

distribution, moisture density relationship, Atterberg’s limit testing, and CBR were first

carried out before the performance of resilient modulus and UCS tests at three different

moisture contents. Extensive statistical analysis was performed to try and establish

correlation equations that could predict the resilient modulus value of a particular soil from

other easy-to-measure index and mechanical properties. Later this investigation was

extended to evaluate the effects of changes in soil moisture as well as assumed default

resilient modulus properties in predicted pavement performance.

The following conclusions are drawn from this study:

• Moisture content variation has a significant effect on the resilient modulus and

shear strength of plastic soils; increasing moisture content results in rapid reduction

in modulus and strength properties. The effects of these variations on nonplastic

soils was not as clearly evident.

 50

• Most of the soil materials tested exhibited highest strength and modulus properties

when tested on the dry side of OMC. Several soils failed to withstand the stress

levels applied during resilient modulus testing, highlighting the importance of

minimizing moisture content fluctuations in the subgrade during the service life of

a pavement.

• Correlation equations developed to predict summary resilient modulus or MEPDG

model parameter values did not have any physical meaning. Although some of the

equations gave reasonably good coefficient of determination (R2) values,

inconsistencies with commonly observed engineering trends means these

equations should not be implemented into practice.

• Total permanent deformation (rutting) was the controlling pavement performance

criterion from pavement analysis and performance prediction using

AASHTOWare Pavement ME Design. Laboratory-established resilient modulus

values were consistently lower than the default values used by ITD; this led to

significant over-prediction in pavement service life when subgrade soil properties

were assumed instead of being established through laboratory testing.

• Linear elastic analysis of typical pavement sections showed that changing the

subgrade resilient modulus primarily affected the vertical (elastic) deformation of

the pavement surface, and can be directly linked to higher rut accumulation in the

subgrade. On the other hand, changes in subgrade resilient modulus did not affect

the horizontal strain at the bottom of the asphalt layer, therefore, not impacting the

fatigue service life of the pavement significantly.

 51

• Findings from finite element analysis also gave evidence in favor of the previous

statement that vertical strain at the top of subgrade changed drastically when

moisture level was changed from OMC to wet of OMC conditions.

References

American Association of State Highway and Transportation Officials (AASHTO).

Mechanistic-empirical pavement design guide–a manual of practice, Interim
Edition, Washington, DC: U.S. 2008.

Barksdale, R. D., J. Alba, N. P. Khosla, R. Kim, P. C. Lambe, & M. S. Rahman. Laboratory

Determination of Resilient Modulus for Flexible Pavement Design. Publication
NCHRP Project 1-28. NCHRP, Transportation Research Board, Washington D.C.,
1997.

Ceylan, H., & K. Gopalakrishnan. Computationally Efficient Surrogate Response Models

for Mechanistic–Empirical Pavement Analysis and Design. Structure and
Infrastructure Engineering, 2011. 7(4): 297-306.

Croney, David, and Paul Croney. The design and performance of road pavements. 1991.

Drumm, E. C., Y. Boateng-Poku & T. Johnson Pierce. Estimation of Subgrade Resilient

Modulus from Standard Tests. Journal of Geotechnical Engineering, 1990. 116(5):
774-789.

El-Badawy, S. M., F. M. Bayomy, M. Santi, & C. W. Clawson. Comparison of Idaho

Pavement Design Procedure with AASHTO 1993 and MEPDG Methods.
In Transportation and Development Institute Congress 2011: Integrated
Transportation and Development for Better Tomorrow, 2011. pp. 586-595.

George, K. P. Resilient Modulus Prediction Employing Soil Index Properties. Publication

FHWA/MS-DOT-RD-04-172. FHWA, U.S. Department of Transportation, 2004.

Guide for Mechanistic–Empirical Design of New and Rehabilitated Pavement Structures.

Final Report, NCHRP Project 1–37A. TRB, National Research Council,
Washington, D.C., March 2004
 52

Heukelom, W., & Klomp, A. Dynamic testing as a means of controlling pavements during
and after construction. In International Conference on the Structural Design of
Asphalt PavementsUniversity of Michigan, Ann Arbor (Vol. 203, No. 1). 1962.

Hossain, Z., M. Zaman, C. Doiron, & P. Solanki. Characterization of Subgrade Resilient

Modulus for Pavement Design. In Geo-Frontiers 2011: Advances in Geotechnical
Engineering, 2011. pp. 4823-4832.

Huang, Y. H. "KENLAYER computer program." Pavement Analysis and Design (1993):

100-167.

Lee, W., N. C. Bohra, A. G. Altschaeffl, and T. D. White. Resilient Modulus of Cohesive

Soils. Journal of Geotechnical and Geoenvironmental Engineering, 1997. 123(2):
131-135.

Majidzadeh, K., F. Bayomy, and S. Khedr. Rutting Evaluation of Subgrade Soils in Ohio.
Transportation Research Record: Journal of the Transportation Research Board,
1978.637: 75–84.

Mallela, J., L. Titus-Glover, B. Bhattacharya, M. Darter, & H. Von Quintus. Idaho

AASHTOWare Pavement ME Design User's Guide, Version 1.1. Publication
FHWA-ID/14-211B. FHWA, U.S. Department of Transportation, 2014.

Mishra, D., Tutumluer, E., & Xiao, Y. (2010). Particle Shape, Type, Amount of Fines and
Moisture Affecting Resilient Modulus Behavior of Unbound Aggregates. In Paving
Materials and Pavement Analysis, 2010. pp. 279-287.

Mishra, D. Aggregate Characteristics Affecting Response and Performance of Unsurfaced

Pavements on Weak Subgrades. Ph.D. Dissertation, University of Illinois at
Urbana-Champaign. 2012.

Nazzal, M. D., L. N. Mohammad, & K. Gaspard. Development of Resilient Modulus

Prediction Models for Louisiana Subgrade Soils. Presented at 87th Annual Meeting
of the Transportation Research Board, Washington, D.C., 2008.
 53

Oh, J. H., E. G. Fernando, C. Holzschuher, & D. Horhota. Comparison of Resilient

Modulus Values for Florida Flexible Mechanistic-Empirical Pavement Design.
International Journal of Pavement Engineering, 2012. 13(5): 472-484.

Rahim, A. M. and K. P. George. Subgrade Soil Index Properties to Estimate Resilient

Modulus. Presented at 83th Annual Meeting of the Transportation Research Board,
Washington, D.C., 2004.

Soliman, H & A. Shalaby, A. Sensitivity of Subgrade Resilient Modulus to Moisture

Variation. In Proceedings from 2010 Annual Conference of the Transportation
Association of Canada, Halifax, Nova Scotia, 2010.

Taylor, A. Mechanistic Characterization of Resilient Moduli for Unbound Pavement Layer

Materials. PhD diss., 2008

Team, R. C. R: A language and environment for statistical computing. R Foundation for

Statistical Computing, Vienna, Austria. 2013.

Thompson, M. R., & Elliott, R. P. ILLI-PAVE based response algorithms for design of
conventional flexible pavements. Transportation Research Record, 1043, 50-57.
1985.

Tutumluer, E. Predicting Behavior of Flexible Pavements with Granular Bases. Ph.D.

Dissertation, Georgia Institute of Technology, Atlanta, Georgia, 1995.

Uzan, J. Influence of the interface condition on stress distribution in a layered system

(abridgement). Transportation Research Record, (616). 1976.

Yau, A., & L. H. Von Quintus. Study Of LTPP Laboratory Resilient Modulus Test Data
and Response Characteristics. Publication FHWA-RD-02-051. FHWA, U.S.
Department of Transportation, 2002.
 54

CHAPTER THREE: MANUSCRIPT TWO – MOISTURE SENSITIVITY OF

TYPICAL IDAHO SUBGRADE SOILS AND THEIR IMPLICATION ON M-E

PAVEMENT DESIGN

Abstract

Subgrade plays an important role in the pavement structure, and mostly contains fine-

grained soils. The moisture sensitivity of subgrade soil has drawn the attention of several

researchers, and has led them to study soil behavior at different moisture contents. This

research study focused on two different soil types collected form two different regions in

the state of Idaho. Special care was taken while selecting the soils to ensure that the soils

were from geographic regions with significantly different annual precipitation trends.

California Bearing Ratio (CBR) and Unconfined Compressive Strength (UCS) tests carried

out on the two soils indicated very high degree of moisture sensitivity, highlighted the

importance of preventing excessive moisture ingress into the subgrade. The recently

developed Mechanistic Empirical Pavement Design method considers resilient response of

soils to predict pavement performance under traffic and environmental loading; this

necessitates determination of resilient modulus of the soils. Both the soils experienced

significant drop in resilient modulus with increasing moisture content, with one of the soils

accumulating more than 5% strain during conditioning cycles. Among different pavement

performance indicators, the M-E design method predicts subgrade rutting using resilient

modulus and several other index properties. Resilient modulus properties for unbound

materials are usually determined in the laboratory using standard test protocols such as
 55

AASHTO T 307; these test protocols only subject the specimen to limited number of

repeated load cycles. This is not representative of the loading scenario in actual pavement

sections, were the subgrade may be subjected to millions of load cycles. To accurately

characterize the permanent deformation potential of soils, a new test protocol was

developed in this study. Stress states to be applied during the permanent deformation

testing were first determined through analysis of typical pavement sections under loading.

Repeated load triaxial tests were performed on the soil specimens by applying the pre-

determined stress levels. Each specimen was tested by applying 10,000 load cycles. Plastic

strain potential of both soils increased significantly when the moisture content was

increased beyond the Optimum Moisture Content (OMC). Soil specimens prepared on the

dry side of OMC, on the other hand, performed significantly better with no permanent

strain accumulation beyond the initial few cycles. Significant change in soil behavior under

loading highlighted the importance of considering different moisture contents and taking

special measures to prevent excessive ingress of moisture into the subgrade during

pavement design and construction, respectively.

Key Words: Subgrade, California Bearing Ratio, Unconfined Compressive

Strength, Mechanistic Empirical Pavement Design, Resilient Modulus, Permanent

Deformation, Moisture Sensitivity

 56

Introduction

Subgrade is considered to be the foundation for a pavement structure, and requires

adequate characterization for the overlying pavement structure to perform satisfactorily

under vehicular and environmental loading. Generally, pavement structures are built on the

existing soil at a site unless special circumstances require its replacement with borrow

material. As subgrade soils are generally fine grained, they are often susceptible to moisture

related failure. Standard construction specifications require the subgrade soil to be

compacted to target density levels (as a percentage of the Maximum Dry Density of MDD)

at or near the Optimum Moisture Content (OMC). The OMC and MDD values are

established through laboratory testing where compaction curves are developed using the

standard or modified compactive efforts. Therefore, most laboratory tests aimed at

establishing strength and modulus properties of subgrade soils use the OMC as the standard

moisture content for specimen preparation. However, establishing all relevant soil

properties at OMC may not present the full picture regarding soil behavior under varying

moisture contents.

The recently developed Mechanistic Empirical (M-E) Pavement Design approach

implemented through AASHTOWare® Pavement ME Design requires subgrade resilient

modulus as the primary input to define its behavior under loading. As the resilient modulus

value is input during M-E pavement design as a single parameter, the test is often carried

out in the laboratory under OMC conditions. The Enhanced-Integrated Climatic Model

(EICM) inherent to Pavement ME Design then modifies the specified subgrade resilient

modulus based on changes in soil moisture and temperature. However, this variation

incorporated through the EICM may not adequately capture the moisture sensitivity
 57

exhibited by subgrade soils. Resilient modulus has been shown by researchers in the past

to exhibit high level of moisture sensitivity (George, 2004). Nazzal et al. (2008) observed

50% to & 70% decrease in resilient modulus with increasing moisture content after

studying four groups of soils at different moisture contents. Subgrade soils of Manitoba,

Canada were tested at wet side of optimum moisture content and it was found that

specimens often accumulated greater than 5% plastic strain during resilient modulus testing

(Soliman and Shalaby 2010). These previous studies indicated the significance of

considering moisture variation during repeated load testing of subgrade soils.

Besides being used to for resilient modulus determination, repeated load triaxial

testing is also used to characterize the permanent deformation properties of unbound

materials. Note that the rutting potential of unbound materials can be evaluated through

this test; rutting is considered to be one of the controlling pavement performance indicators

during M-E design. Note that this is more significant for flexible pavement structures

compared to rigid pavements because of the difference in which loads are dissipated in the

two pavement types. Excessive subgrade rutting can significantly reduce the service life of

a pavement (Monismith et al. 1972; Barksdale 1972). Majidzadeh et al. (1978) reported

that weak subgrade soils were generally susceptible to rutting, and can contribute more

than 40% towards total pavement rutting. The M-E pavement design approach currently

uses soil resilient modulus and index properties to make predictions regarding the

permanent deformation of soils; this may not represent the best approach because resilient

modulus is an indicator of the “elastic” behavior of a soil, whereas permanent deformation

accumulation is a “plastic” phenomenon. Puppala et al (1996) found that fine grained soils

such as silty clay and silty sand showed good resilient modulus, but experienced significant
 58

rutting under traffic loading during their service lives. It is therefore necessary to

characterize both resilient as well as plastic deformation properties of subgrades to

facilitate the design and construction of well-performing pavement structures. Adequate

information concerning the permanent deformation behavior of soils can help in the

determination of suitable pavement layer thicknesses to minimize excessive rut

accumulation under loading (Puppala et al, 1999).

A research study was recently undertaken in collaboration with the Idaho

Transportation Department (ITD) to evaluate the permanent deformation characteristics of

some of the common soil types used in pavement applications across the state of Idaho.

One of the primary goals was to evaluate the moisture sensitivity of these soils, and

quantify how the resilient modulus and permanent deformation properties change with

changes in moisture content. These investigations would help transportation agencies to

make important decisions for implementing M-E pavement design practices, and would

encourage them to develop state-specific guidelines and practices to restrict rut

accumulation in the subgrade within allowable limits. Tasks undertaken under the scope of

this research effort to accomplish the overall objective have been listed:

1) Identify regions in the state of Idaho that experience significantly different

precipitation amounts throughout the year, and collect soils from these two regions;

2) Conduct laboratory tests to characterize the resilient modulus, shear strength, and

permanent deformation properties, and study the effect of moisture content

variation on these properties;

 59

3) Develop a new testing protocol to characterize the permanent deformation behavior

of fine-grained soils. Use this new test protocol to study the effect of moisture

content variation on the susceptibility of fine-grained soils to rutting.

Material Selection

As already mentioned, the primary objective during the material selection phase

was to identify two locations in the state of Idaho that receive significantly different annual

precipitations, and collect soil samples to study how the differences in soil moisture content

may affect pavement performance. A precipitation map for Idaho was prepared to aid

identification of the candidate locations for soil sampling. Precipitation data was collected

from Idaho State Climate Service (ISCS) and then the annual average precipitations were

plotted on a map (see Figure 3-1) using the ArcGIS software.

 60

Figure 3-1 Annual average precipitation map for Idaho

The state of Idaho has a widely varying geography and weather where some regions

are very dry, and some regions experience significantly higher amounts of rainfall. As seen

from Figure 3-1, the northern part of the state receive significantly higher precipitation

compared to the southern parts. One soil (termed S1) from the northern part of Idaho was

selected for this study to represent ‘wet zone’ conditions. Another soil (termed S2) was

collected from the south-west region of the state, and represented the ‘dry zone’ conditions.
 61

Preliminary Soil Characterization and Shear Strength Measurements

Grain Size Distribution

The grain size distribution for the two soils was established using the AASHTO T

88 test method, and have been plotted in Figure 3-2. The amount of sand, silt and clay

particles of these soils have been listed in Table 3-1. Atterberg’s limit tests were carried

out per AASHTO T 90 standard; both soils exhibited low plasticity index (PI), and could

be classified as ‘A-4’ according to the AASHTO classification system. The equivalent

USCS classifications were: CL for S1 and CL-ML for S2.

Figure 3-2 Grain size distribution of selected two soil types

 62

Table 3-1 Soil Index Properties and Classification

Material Liquid Plasticity Sand Silt Clay Soil Classification

Limit Index
LL PI

% % % % % USCS AASHTO

S1 30 3 51.2 43.7 5.1 CL A-4

S2 26 5 46.1 47.2 6.7 CL-ML A-4

Moisture Density Relationship

The optimum moisture content (OMC) and maximum dry density (MDD) of the

two soils were established in the laboratory following the AASHTO T 99 test method. This

information was subsequently used in this study to prepare specimens for resilient modulus,

shear strength, and permanent deformation testing. Figure 3-3 shows the moisture-density

plots for the two soil types. As expected, the dry density increased with the increasing

moisture content until OMC, and decreased beyond that. To study the effect of moisture

content variation on soil behavior, specimens were also prepared in this study at 90% OMC

(dry) and 110% OMC (wet) conditions. Table 3-2 lists the moisture contents for the two

soils corresponding to the dry, OMC, and wet conditions.

 63

Figure 3-3 Compaction moisture-density characteristics for the two soils

established following the AASHTO T 99 procedure

Table 3-2 Moisture Contents Corresponding to Dry, OMC, and Wet Conditions
for the Two Soil Types

Soil ID Compaction Moisture Content

Dry OMC Wet

S1 15.1 16.8 18.5

S2 17.8 16.2 14.6

Strength Properties of Soils for Typical Pavement Design

California Bearing Ratio (CBR) and Unconfined Compressive Strength (UCS)

were used in the current study to determine establish shear strength properties for the two

soil types. Although the CBR is an index property, it has been used extensively by

researchers and practitioners as an indicator of unbound material shear strength. CBR is

 64

also used as an indirect method to predict the resilient modulus of soils (Patel et al 2010).

Several different correlations equations have been used by state and local highway agencies

to predict resilient modulus from CBR and UCS value (Jackson 2015); the resilient

modulus values thus predicted, are used as inputs during M-E pavement design instead of

having to perform complex and cumbersome repeated load triaxial testing. CBR values for

the two soil materials were determined in the current study under unsoaked conditions, but

otherwise following the AASHTO T193 method. Each soil material was tested at three

different moisture conditions to quantify the moisture sensitivity; CBR values for the two

soils under the three different moisture conditions have been listed in Table 3-3. The same

data has been plotted in the form of a column plot in Figure 3-4 (a). As seen from Table 3-

3 as well as Figure 3-4 (a), both soil types showed significant degree moisture sensitivity.

Under dry conditions, both soils exhibited CBR values that were at least four times higher

than the corresponding values under OMC conditions. On the other hand, when the

moisture content was increased to 110% OMC, the CBR value underwent a rapid reduction.

Soil S1 exhibited a greater level of moisture sensitivity compared to S2 during CBR testing.

Table 3-3 California Bearing Ratio and Unconfined Compressive Strength of

Soils

Soil Type Conditions CBR (%) UCS (kPa)

Dry 24 203

S1 OMC 6 155

Wet 2 121

Dry 17 221

S2 OMC 4 166

Wet 2 110
 65

Unconfined compressive strength values for the soils were established following

guidelines provided in AASHTO T208. After completion of the resilient modulus tests

(discussed in the next section), the same specimen was used to test for UCS under

controlled strain conditions at a rate of 1 mm/min. Similar to the CBR testing, specimens

were prepared at three different moisture conditions and tested to quantify their moisture

sensitivity. UCS values for the two soils under three different moisture conditions were

also listed in Table 3-3, and the same data has been plotted in the form of column plot in

Figure 3-4 (a). As seen from the table and the plot, both soil types exhibited highest UCS

values when tested under dry conditions. UCS values decreased significantly as the

moisture contents were increased to OMC and subsequently to wet of OMC conditions.

(a) (b)

Figure 3-4 (a) California Bearing Ratio and (b) Unconfined Compressive
Strength of soils at different conditions
 66

Resilient Response for Pavement ME Design

Resilient modulus properties of the two soils were established in the laboratory using

the AASHTO T 307 test protocol. Besides providing desired information for M-E

pavement design, the resilient modulus data was also used during pavement analyses to

calculate the stress states on top of the subgrade, which would then be used to develop the

new permanent deformation test protocol. Resilient modulus characterization is

representative of pavement layer response under stress levels that are significantly lower

than the corresponding shear strength values which is necessary to model the performance

of pavement layers (Taylor 2008). Resilient modulus testing was conducted at three

different moisture contents to study the effects of varying moisture contents. Two replicate

specimens were tested for each soil type at the three different moisture contents. Figure 3-

5 shows the variation in resilient modulus of the two soils due to changing moisture

content. For the soil S1, the resilient modulus value decreases consistently with increasing

moisture content; this effect is more prominent at high confining stress levels. At low

confining stress levels, although the transition from ‘dry’ to OMC conditions does not

appear to have a drastic effect, there is a sudden drop in MR observed when the moisture

changes from OMC to ‘wet’. Note that the confining pressures felt by a subgrade soil

element underneath a typical pavement structure is usually very low (of the order of 7-14

kPa). Accordingly, resilient modulus values for the soils at low confining pressure levels

would be representative of field conditions.

 67

(a) (b)

Figure 3-5 Resilient Modulus values for the two soils under different moisture
conditions: (a) soil type S1 and (b) soil type S2

On the other hand, soil S2 experienced greater than 5% permanent strain

accumulation during the conditioning phase under wet conditions; this is taken to be an

indicator of ‘specimen failure’ during MR testing, therefore rendering this soil unsuitable

for pavement applications under wet conditions. It should be noted that under no stress

level, did the dry specimens of S2 exhibit significantly stiffer behavior compared to the

specimens tested under OMC conditions. There are several models available to predict the

resilient modulus of soils based on different stress states. These model parameters are

required during the pavement analysis to compute the stress-strain experienced at different

points within the pavement structure under traffic loading. In this study, the MEPDG model

and Uzan model were chosen and fitted to the laboratory test data; the model parameters

established for the two soil types have been listed in Table 3-4. The two model forms have

been given below.

 68

k2 k3
 θ  τ 
M R = k1 × pa ×   ×  oct + 1 MEPDG Model
 pa   pa 

MR =k × (θ ) × (σ d )
n m
Uzan Model

where
θ is known as bulk stress or the first stress invariant (θ = σ 1 + σ 2 + σ 3 )
σ 1 is the major principal stress
σ 2 is the intermediate principal stress
σ 3 is the minor principal stress
σ d is the deviator stress
For a cylindrical triaxial test set-up
σ2 = σ3
σ=
1 σ3 +σd
σ1 − σ 3 =
σd
pa is atmospheric pressure, and equals 101.3 kPa
The pa term is used for normalization in the equation
τ oct represents the octahedral shear stress
1
τ oct= (σ 1 − σ 2 ) + (σ 2 − σ 3 ) + (σ 3 − σ 1 )
2 2 2

3
For a cylindrical triaxial test specimen,
2 2
τ oct = (σ 1 − σ 3 ) = × σ d
3 3
 69

Table 3-4 MEPDG and Uzan Model Parameters of Soils at Different Moisture
Conditions

MEPDG Model Parameters Uzan Model

Soil
Condition
Type
k1 k2 k3 R2 k n m R2

Dry 788.4 0.82 -3.11 0.88 1466.75 0.84 -0.44 0.95

S1 OMC 637.21 0.24 -1.73 0.75 5354.27 0.25 -0.25 0.89

Wet 164.54 0.62 0.54 0.82 440.86 0.63 0.06 0.81

Dry 555.85 0.57 -2.66 0.86 1997.5 0.58 -0.36 0.91

S2 OMC 377.2 0.68 -2.68 0.76 1025.7 0.68 -0.37 0.81

Wet - - - - - - - -

As seen from the above table, both k3 and m parameter are negative for dry and

OMC condition, but it becomes reverse at wet condition. This means that response of soils

due to deviator stress depends on moisture condition. The Uzan model was used in this

study to analyze the response of typical pavement structures under vehicular loading; this

information was then used to develop the new permanent deformation test protocol as

discussed in the next section.

Finite Element Analysis of Typical Pavement Sections

Tensile strain at bottom of HMA layer, as well as vertical compressive strain and

deviator stress at the top of subgrade layer are considered as the critical pavement responses

which are often used to predict the service life of pavements. In this section, the effect of

subgrade moisture condition was evaluated on pavement response by monitoring changes

in the critical pavement response parameters. It was decided to use ILLI-PAVE 2005

software to compute the critical pavement responses as ILLI-PAVE considers the stress

dependent nature of unbound materials (Thompson & Elliott 1985). Table 3-5 lists the
 70

material properties that were used as inputs in ILLI-PAVE. The HMA and base layer

properties were taken from the Idaho AASHTOWare® Pavement ME design user guide

(Mallela et al, 2018). The subgrade properties were assigned based on values determined

in the laboratory for S1 and S2. Soil properties for all three moisture conditions (dry, OMC,

and wet) were used to analyze different pavement sections. Finally, three different

pavement sections with different layer thickness were considered to cover the typical range

of pavement structures constructed in the state of Idaho. Note that specimens of soil S2

prepared under wet conditions were unable to sustain the stress levels applied during

repeated load triaxial testing, and therefore, that particular case has not been included in

this analysis. Figure 3- 6 shows the critical pavement response parameters obtained upon

subjecting the pavement sections to a 40-kN wheel load at 80 psi tire pressure.

Table 3-5 Summary of Material Properties for ILLI-PAVE Analysis

Density Modulus Poisson’s Layer Thickness (mm)

Materials 3
(kg/m ) (MPa) Ratio
Sec-1 Sec-2 Sec-3

HMA 2323 3447 0.35 101.6 152.4 203.2

K-θ
Base 2274 0.35 304.8 457.2 609.6
model

Dry 1730

S1 OMC 1730
Subgrade
Soils
Wet 1698
Uzan
model
0.40 ∞ ∞ ∞
Dry 1762
S2
OMC 1762
 71

(a) (b)

(c)

Figure 3-6 Tensile strain at bottom of HMA layer (b) Compressive strain at top
of subgrade (c) Deviator stress at top of subgrade

As seen from the figure, under OMC conditions, soil S1 exhibits lowest tensile

strain at the bottom of HMA layer but the same response parameters is lower for soil S2

under dry conditions. Vertical Strain on top of subgrade becomes worse when moisture

content changes from OMC to wet conditions for soil S1; the value is much lower under

OMC conditions. However, for soil S2, the response under dry conditions is better

compared to OMC. Deviator stress on top of the subgrade is governed by stiffness of the

subgrade layer, which is significantly influenced by the moisture content. High amount of

moisture makes a soil weaker, and that leads greater settlements (or strains) and lower
 72

stresses (as stress = stress x modulus). As shown from the plots, lower deviator stresses on

top of the subgrade were obtained for soil S1 under wet conditions. Also, this value was

not found to change significantly for either soil when the moisture level was changed from

dry to OMC.

Laboratory Tests to Characterize the Permanent Deformation Potential of Soils

PD test program

As already mentioned, no standard specification is available to characterize the

permanent deformation (PD; or rutting) potential of unbound materials in the laboratory.

Accordingly, this study attempted to develop a new test protocol that would consider the

typical stress states experienced by soils in a pavement structure rather than using some

pre-established stress levels. Several researchers have recently (Rahman et al. 2019)) used

data from resilient modulus tests to evaluate the permanent deformation potential of

unbound materials. However, it should be noted that the MR test procedure applies only

around 2500 cycles of loading to the specimen, which is not sufficient to study the PD

characteristics of soils and aggregates. Moreover, the MR test procedure (per AASHTO

T307) applies 15 different stress states sequentially to the specimen. This creates stress-

history effects on the specimen behavior, and therefore, can be misleading if used for PD

characterization. Finally, some of the stress levels applied during the MR testing are higher

than those experienced by soil elements underneath typical pavement structures.

Considering the above-mentioned uncertainties, the current study selected 41.4 kPa

and 55.2 kPa deviator stress levels based on results from ILLI-PAVE analysis (shown in

Figure 3-6c). Both the soils were tested under two different deviator stress levels to

characterize the permanent deformation response. Generally, confining pressure levels in

 73

subgrade are very low, and can often be lower than 34.5 kPa (Thompson and Robnett

1976). The current study chose to perform the PD tests at a confining pressure level of 20.7

kPa. The soil specimens were 100 mm in diameter, and 200-mm tall. A standard

compaction drop hammer was used to compact the specimen in three layers targeting 100%

maximum dry density. Moreover, adhesion between the layers was ensured by scarifying

the surface of the compacted layer before placement of the next layer. Specimens were

prepared under moisture contents corresponding to dry, OMC, and wet conditions. The

deviator stress levels calculated from ILLI-PAVE analyses were applied to the specimen;

haversine load pulses were applied with a 0.1s loading period followed by a 0.9s rest

period. Each specimen was subjected to 10,000 cycles of loading; axial strain data was

recorded during the testing using two externally mounted LVDTs on top of the triaxial cell.

Two replicate specimens were tested at each soil-deviator stress-moisture condition

combination to confirm repeatability of the test results.

Analysis of PD Test Results

After completion of the PD tests, axial permanent (plastic) strain data was plotted

against the number of loading cycles (see Figure 3-7 for S1 and Figure 3-8 for S2).

Accumulated plastic strain under repeated loading is representative of the susceptibility of

soils to rutting under traffic loading. Under the first few load cycles, the permanent strain

accumulation increases rapidly, with the rate of permanent strain accumulation reducing

significantly as the specimen approaches “stable” conditions; ultimately, the specimen

attains hysteretic stress-strain response under repeated loading.

 74

(a) (b)

Figure 3-7 Accumulation of plastic strain in soil S1 when tested at two different
deviator stress values: (a) σ d = 41.4 kPa ; and (b) σ d = 55.2 kPa

As seen from the above figure, soil S1 under dry and OMC conditions became

stabilized after only 200 cycles of loading; the S1 specimen under wet conditions took 1000

cycles to reach ‘stable’ behavior (characterized by no significant permanent strain

accumulation under further loading). The figure also shows that plastic strain accumulation

after 10,000 cycle under wet conditions is greater than that for dry and OMC conditions.

Note that the S1 specimen under wet conditions accumulated plastic strains that were 8

times greater than that under OMC conditions at end of testing. S1 soil specimens at both

dry and OMC conditions exhibited relatively low plastic strain accumulation upon 10,000

cycles of loading; this was consistent for both the deviator stress levels. Note that soil S1

was collected from the northern part of Idaho form a region that experiences significantly

high amounts of annual precipitation. This soil also exhibited very high susceptibility

towards rutting under wet conditions. It is therefore recommended that during the design

and construction of pavements in this region, extra care should be taken to prevent the

ingress of water into the subgrade layer. Moreover, the design should be checked by using
 75

lower subgrade modulus values (to represent wet conditions) to ensure that the pavement

layer thicknesses are adequate to protect the subgrade from excessive rutting.

Figure 3-8 shows similar plots for the S2. Note that S2 exhibited very high levels of

moisture sensitivity, and the specimen prepared under wet conditions failed during PD

testing when the deviator stress level was equal to 55.2 kPa. Note that when subjected to a

deviator stress level of 41.4 kPa, the S2 specimen at wet condition accumulated more than

3% plastic strain only after 500 load cycles. Unlike the S1 specimen, specimens for S2

exhibited high sensitivities even to changes in deviator stress magnitudes. Figure 3-8 shows

that plastic strain accumulation rate under OMC conditions were significantly different

from those under dry conditions. Under dry conditions, the specimens reached ‘stable’

behavior after only 100 cycles. As the S2 material was collected from a region of Idaho

with relatively low precipitation, designing for OMC conditions may be appropriate in this

case.

(a) (b)

Figure 3-8 Accumulation of plastic strain in soil S2 when tested at (a) 41.4 kPa
deviator stress (b) 55.2 kPa deviator stress
 76

Summary and Conclusions

This chapter presented results from a laboratory testing effort targeted at characterizing

the permanent deformation potential for two different subgrade soil types under different

moisture conditions. Laboratory tests performed on the soils included: (1) gradation; (2)

Atterberg’s limits; (3) moisture-density characteristics; (4) resilient modulus; (5)

unconfined compressive strength; and (6) permanent deformation. The modulus, shear

strength, and permanent deformation tests were carried out under three different moisture

conditions. The resilient modulus test data were fitted with MEPDG, Uzan models to

establish the model parameters, which were then used during finite element analysis of the

pavement sections to calculate critical pavement response parameters under traffic loading.

The analysis was performed using ILLI-PAVE as it can adequately capture the stress-

dependent behavior of unbound materials. Stress states experienced by the subgrade layer

under typical pavement configurations were calculated corresponding to the three different

subgrade moisture conditions. These stress states were then used to develop a new

permanent deformation test protocol, and the effects of moisture content and stress levels

on permanent deformation behavior were studied. The following important conclusions

can be drawn from this study:

(1) Both soils exhibited high levels of moisture sensitivity. CBR, UCS, as well as MR

values decreased significantly as the moisture content was increased. The change in

these values was worse when the moisture content was increased from OMC to 110%

OMC compared to the change from 90% OMC to OMC.

 77

(2) Pavement response calculations under standard wheel loading indicated that

vertical strain on top of the subgrade increased and the deviator stress decreased under

wet conditions.

(3) Permanent deformation testing of the soils showed that both soils can undergo

significant rutting under wet conditions. Both soils may reach ‘stable behavior’ under

construction traffic if dry conditions are maintained by ensuring proper drainage.

However, significant rutting may result if the moisture content increased during the

service life of the pavement. New pavement design or rehabilitation projects in regions

experiencing high amounts of precipitation should consider these factors and need to

ensure proper drainage of pavement layer to protect the subgrade layer. Using subgrade

modulus values at OMC conditions for all designs is not the best approach at locations

where significant fluctuations in subgrade moisture content is likely.

Limitations of Current Study and Recommendations for Future Research

Although this research effort successfully accomplished many of its objectives, several

limitations associated with the testing effort need to be considered. These limitations are

listed below, and should be considered during future research efforts.

(1) The soil specimens in this study were tested for PD characteristics under the

application of 10,000 load cycles, which is which is well below the number of

traffic load repetitions experienced by pavement layers throughout their lifetime. A

more extensive PD study needs to test the specimens up to 100,000 load

applications.

(2) In this study, soils were only tested at two different deviator stress levels. This is

not sufficient to develop an extensive rutting prediction model. The

 78

AASHTOWare® Pavement ME Design software uses the MEPDG rutting model

which predicts unbound material rutting based on resilient modulus and other soil

index properties. Adequacy of the MEPDG model needs to be evaluated by

comparing model-predicted rutting with actual lab tested rutting values for

subgrade soils. However, such an effort will require the testing of several soil types

and testing under several different stress states. As only two soils were tested in this

effort, and due to time and material availability related constraints, the current study

could not undertake the effort to calibrate the MEPDG model for Idaho conditions.

This will be addressed in future research efforts.

References

Barksdale, R. D. Laboratory evaluation of rutting in base course materials. In Presented at

the Third International Conference on the Structural Design of Asphalt Pavements,
Grosvenor House, Park Lane, London, England, Sept. 11-15, 1972. (Vol. 1, No.
Proceeding), 1972.

George, K. P. Resilient Modulus Prediction Employing Soil Index Properties. Publication

FHWA/MS-DOT-RD-04-172. FHWA, U.S. Department of Transportation, 2004

Jackson, K. D. Laboratory resilient modulus measurements of aggregate base materials in

Utah, 2015.

Majidzadeh, K., Bayomy, F., & Khedr, S. Rutting evaluation of subgrade soils in Ohio,
1978.

Mallela, J., Titus-Glover, L., Bhattacharya, B., Darter, M., & Von Quintus, H. Idaho
AASHTOWare pavement ME design user's guide, version 1.1 (No. FHWA-ID/14-
211B). Idaho. Transportation Dept., 2014.

Monismith, C. L., Ogawa, N., & Freeme, C. R. Permanent deformation characteristics of

subgrade soils due to repeated loading. Transportation Research Record, 537,
1975.
 79

Nazzal, M. D., L. N. Mohammad, & K. Gaspard. Development of Resilient Modulus

Prediction Models for Louisiana Subgrade Soils. Presented at 87th Annual Meeting
of the Transportation Research Board, Washington, D.C., 2008.

National Cooperative Highway Research Program. "Harmonized test methods for

laboratory determination of resilient modulus for flexible pavement design." Final
Rep. NCHR Project No. 1-28 A (2003).

Patel, R. S., & Desai, M. D. CBR predicted by index properties for alluvial soils of South
Gujarat In Proceedings of the Indian geotechnical conference, Mumbai (pp. 79-82),
2010.

Puppala, A. J., Mohammad, L. N., & Allen, A. Engineering behavior of lime-treated

Louisiana subgrade soil. Transportation Research Record, 1546(1), 24-31, 1996

Puppala, A. J., Mohammad, L. N., & Allen, A. Permanent deformation characterization

of subgrade soils from RLT test. Journal of Materials in Civil Engineering, 11(4),
274-282, 1999.

Rahman, M. M., K. M. Islam, S. L. Gassman. Comparative Study of Permanent Strain and

Damping Characteristics of Coarse Grained Subgrade Soils with Resilient Modulus
Presented at 98th Annual Meeting of the Transportation Research Board,
Washington, D.C., 2019.

Soliman, H & A. Shalaby, A. Sensitivity of Subgrade Resilient Modulus to Moisture

Variation In Proceedings from 2010 Annual Conference of the Transportation
Association of Canada, Halifax, Nova Scotia, 2010.

Taylor, A. Mechanistic characterization of resilient moduli for unbound pavement layer

materials (Doctoral dissertation), 2008.

Thompson, M. R., & Robnett, Q. L. Resilient properties of subgrade soils (No. FHWA-
IL-UI-160 Final Rpt.), 1976.

Thompson, M. R., & Elliott, R. P. ILLI-PAVE based response algorithms for design of
conventional flexible pavements. Transportation Research Record, 1043, 50-57.
1985.
 80

CHAPTER FOUR: SUMMARY AND CONCLUSIONS

Summary

The primary focus of this thesis was to characterize different subgrade soil types

collected from across the state of Idaho through resilient modulus and permanent

deformation testing along with other conventional, commonly used laboratory tests.

Preliminary tests carried out in the laboratory included sieve analysis, Atterberg’s limit

testing and moisture-density characterization. Moreover, other commonly used strength

and index tests such as Unconfined Compressive Strength (UCS) and California Bearing

Ratio (CBR) were also carried out. Repeated load triaxial testing was carried out according

to the AASHTO T307 protocol to measure resilient properties of the subgrade soils.

Specimens were prepared and tested at three different moisture contents (dry of optimum,

optimum, and wet of optimum) to quantify the effects of moisture variation on resilient

modulus. The conventional stress-dependent resilient modulus models such as the ones

proposed by Uzan, and the one used in the MEPDG (known as the universal model) were

used to fit the laboratory data; the corresponding model parameters were thoroughly

analyzed. The resilient modulus test results were analyzed along with the UCS and CBR

test results to investigate the feasibility of predicting the resilient modulus values of

subgrade soils without the need to run the complex repeated load triaxial tests. The effect

of assumed resilient modulus and moisture variation in soils were investigated by

computing pavement response and predicted pavement performances.

 81

Stress levels on top of the subgrade under standard wheel loads were estimated

through finite element analysis of typical pavement sections by considering stress-

dependent unbound material behavior. This information was later used to develop a new

permanent deformation (PD) testing protocol. PD tests were conducted on soil specimens

prepared at different moisture contents to study moisture sensitivity of plastic strain

accumulation in soils.

Conclusions

The following conclusions are drawn from this study:

• Fine grained soils showed high level of moisture sensitivity with little variation in

moisture content; deviator stress levels had varied impact on MR based on whether

the soil comprised non-plastic or plastic fine.

• The reduction in soil modulus and strength properties when moisture content is

increased from dry to optimum conditions is not as drastic as when the moisture

content is increased beyond optimum condition.

• Although some correlation equations can be developed to predict the resilient

modulus value of subgrade soils from easy-to-establish index and mechanical

properties, most of these equations do not match commonly-observed material

behavior trends. Therefore, generic equations to predict resilient modulus from

other soil properties can lead to erroneous pavement designs.

• Subgrade soils exhibit very high levels of moisture sensitivity. Ultimate care needs

to be taken during pavement construction to ensure the subgrade layer is

compacted under dry conditions; adequate measures should be taken to prevent

ingress of excessive moisture into the pavement structure during its service life.
 82

• Even if some soils exhibit reasonable resilient modulus values when tested at OMC

or wet of OMC conditions, they may accumulate excessive permanent deformation

when subjected to very high number of load applications. Therefore, predicting

rutting potential of soils using resilient modulus is not a recommended approach.

• Moisture sensitivity of soils should be considered during the design and

construction of pavement sections in regions that experience very high amounts of

precipitation. Using the soil properties determined at optimum moisture conditions

for all design types is not recommended.