Downloaded from ascelibrary.org by UNIVERSITY OF FLORIDA on 11/10/15. Copyright ASCE. For personal use only; all rights reserved.

Advanced Characterization of Granular Materials

for Mechanistic Based Pavement Design

Umit Seyhan 1, Student Member and Erol Tutumluer2, Member

Abstract

The recent adoption and use of mechanistic concepts in pavement design

made it possible to more realistically characterize the unbound aggregates
component of the pavement by incorporating the new advances in characterization,
i.e., anisotropic, stress path dependent behavior, and accounting for the most
damaging field conditions. This paper mainly focuses on the use of a unique
triaxial testing machine, referred to as University of Illinois FastCell (UI-FastCell),
for determining in the laboratory the stress-induced anisotropic resilient properties
of twelve aggregates with varying material types and properties. In the selection
process, consideration was given to both "good" and "poor" performing granular
base/subbase materials obtained from seven different states. With the applied
AASHTO T294-94 stress states conveniently pulsed either in the vertical or
horizontal direction on the same specimen, the anisotropy (directional dependency)
of resilient moduli are adequately determined in the laboratory for the compression
and extension type dynamic loadings. Nonlinear stress dependent models are
developed to characterize the resilient moduli under these extreme loadings. The
moduli obtained under extension stress states are in general lower than the ones
computed from compression tests. Such lower moduli would result in higher
critical pavement responses as obtained from the mechanistic analysis of a
conventional flexible pavement and be associated with the reduced pavement life
eventually. The most significant stress history effect is observed for the materials
having moderate amount of fines. Finally, anisotropic (horizontal to vertical)
modular ratios corresponding to horizontal (extension) and vertical pulsing
(compression) conditions are shown to either increase or decrease with increasing
applied dynamic stresses depending on the material properties, i.e., quality of an
aggregate.

' Graduate Research Assistant

2 Assistant Professor
Department of Civil and Environmental Engineering
University of Illinois, 205 N. Mathews, Urbana, 1L 61801.

51

Pavement Subgrade, Unbound Materials, and Nondestructive Testing

 52 PAVEMENT SUBGRADEAND NONDESTRUCTIVETESTING

Introduction
Downloaded from ascelibrary.org by UNIVERSITY OF FLORIDA on 11/10/15. Copyright ASCE. For personal use only; all rights reserved.

The principal design procedure for highways is the AASHTO Design

Guide, which is currently being updated for publication in 2002 to include
mechanistic concepts in design. With increasing demands being placed on
highways and airport pavements through heavier and increasing number of loads it
is critical that we be able to better characterize the unbound aggregates component
of the pavement by incorporating the recent advances in our characterization, i.e.,
anisotropic, stress path dependent behavior. In laboratory testing, application of
the actual field stress states on the specimen is essential to better understand and
model the structural behavior of unbound aggregate bases. A new laboratory test
method recently developed to determine stress-induced anisotropic aggregate
stiffnesses has been put to use in this study. The use of such advanced methods
that more realistically characterize flexible pavements having unbound aggregate
bases could significantly improve the ability to reliably predict unstabilized
pavement response, which in turn would lead to better design methodology.

Contrary to the commonly used assumption of isotropy, granular

bases/subbases have anisotropic material properties. This is primarily due to the
effects of anisotropic loading in the field, i.e., initial compaction and the
subsequent wheel loading, the deformability of aggregate layer decreases and a
higher modulus is usually attained in the vertical direction. During these
construction and initial loading stages, horizontal residual stresses, typically
compressive in nature, are locked in the granular layer. In laboratory triaxial
testing, it is important to determine specimen lateral deformations and
independently pulse radial stresses in order to properly establish anisotropic
resilient properties and characterize aggregate response to various compression and
extension type stress states often experienced under the moving wheel loads.

This paper mainly focuses on the use of a unique triaxial testing machine,
referred to as University of Illinois FastCell (UI-FastCell), for determining in the
laboratory the stress-induced anisotropic resilient properties of 12 aggregates,
which have been also studied in the NCHRP 4-23 project. With the applied
AASHTO T294-94 stress states conveniently switched on the same specimen, the
directional dependency of resilient properties are adequately determined in the
laboratory for the compression and extension type dynamic loadings. Nonlinear
stress dependent models are developed to characterize the resilient moduli of the 12
aggregates under these extreme loadings. Finally, the effect of aggregate quality on
anisotropic (horizontal to vertical) modular ratios corresponding to horizontal and
vertical pulsing conditions is investigated.

Stress States In Pavement Granular Layers

To better characterize granular material behavior, it is important to properly

simulate in the laboratory the actual loading conditions. The pavement in the field

Pavement Subgrade, Unbound Materials, and Nondestructive Testing

 PAVEMENT SUBGRADEAND NONDESTRUCTIVETESTING 53

is usually loaded by moving wheel loads, which at any time impose varying
Downloaded from ascelibrary.org by UNIVERSITY OF FLORIDA on 11/10/15. Copyright ASCE. For personal use only; all rights reserved.

magnitudes of vertical, horizontal, and shear stresses in the aggregate layer

accompanied by the rotation of the principal stresses. This type of loading can not
be ideally simulated in the laboratory by the constant confining pressure (CCP)
type repeated load triaxial tests, which have been commonly used in the U.S. since
late 1960s and recognized as the standard procedure (AASHTO T294-94). It is
only possible in the CCP tests to apply one constant stress path (Aq/Ap = 3)
representing those stress states that occur directly under the wheel loading (see
Figure I a).

The variable confining pressure (VCP) type repeated load triaxial tests, on
the other hand, offer the capability to apply a wide combination of stress paths by
pulsing both cell pressure and vertical deviator stress corresponding to the
pulsations of the different magnitudes of horizontal and vertical stresses due to a
passing wheel load. Such stress path loading tests better simulate actual field
conditions since in the pavement structure the confining stresses acting on the
material are also cyclic in nature (see Figure lb). Typically, some radial distance
away from the centerline of loading, the horizontal component of the dynamic
wheel load can become greater in magnitude than the vertical component. In that
case, an extension type of loading can be more critical on top of the base as
illustrated in Figure lb.

(a) Rotation of principal stress axes (b) Stresses in extension loading

Figure 1. Stress Conditions in A Granular Base to Consider for Advanced

Aggregate Characterization

Tutumluer and Seyhan (1999a) recently analyzed a conventional flexible

pavement to study typical q-p stress path slopes and the corresponding principal
stress rotations throughout the base layer. As obtained from finite element

Pavement Subgrade, Unbound Materials, and Nondestructive Testing

 54 PAVEMENT SUBGRADEAND NONDESTRUCTIVETESTING

predictions, triaxial extension states primarily occurred in the granular layer when a
Downloaded from ascelibrary.org by UNIVERSITY OF FLORIDA on 11/10/15. Copyright ASCE. For personal use only; all rights reserved.

compressive residual stress of 21-kPa (3-psi) was included in the analysis as the
initial stress state (Tutumluer and Seyhan, 1999a). Such magnitudes of residual
stresses were reported previously to exist in the compacted granular layers, which
offset any low magnitude horizontal tensile stresses and provided adequate
confinement radially away from the wheel load (Uzan, 1985; Garg et al., 1998).

Due to the moving nature of the wheel, the cyclic component of the load
imposes a change (increment) of stress state, which is typically not co-axial with
the stress state under the static overburden load (see Figure I a). In other words, the
major principal stress due to wheel load is not aligned in the vertical direction, but
rotates in the direction of the applied load as shown in Figure la. Depending upon
the location, the total principal stresses on a pavement element "A", therefore,
rotate by an angle a as the wheel load passes (see Figure 1a).

In a consistent laboratory approach for investigating the anisotropy and the

stress path dependency of the material behavior, it is therefore important to
measure both axial and radial deformations of a repeated load triaxial specimen and
to individually account for the resilient response of the granular material to both
radial and vertical pulse loadings.

Universi~. of Illinois FastCell and Recent Findings

Based on the latest research findings (Tutumluer, 1995; Tutumluer and

Thompson, 1997a-b), a new innovative laboratory testing device, named University
of Illinois FastCell (UI-FastCell), was recently custom-designed and manufactured.
UI-FastCell has provisions for measurement of on sample vertical and radial
displacements as well as pulsing of the major principal stress in the vertical or
radial direction by the use of the two independently controlled stress channels.
Since it is not possible to reorient the granular samples in the triaxial cell, applying
and switching of the various stress states on the same specimen facilitates
determining especially the load-induced anisotropy. The device is also suitable for
simulating field stress conditions in the laboratory and for studying the effects of
principal stress rotation due to moving wheel loads that involve a change in total
shear stress direction.

UI-FastCell uses a fluid/air interface to minimize compressibility effects

when conducting tests in which the horizontal stress on a specimen must be cycled.
This is useful for investigating anisotropic effects and the response to loading in
which a 90 ~ rotation of planes of principal stress is important. Figure 2a shows a
picture of the UI-FC with the confinement cell lowered down on the specimen for
the testing position. An air actuator applies the axial pressure and the confining
pressures are cycled through a hydraulic fluid within the rubber membrane. The
driving cylinders on the back of the confining cell (not shown here) include an air-
fluid interface, which provides fast application and switching of the dynamic

Pavement Subgrade, Unbound Materials, and Nondestructive Testing

 PAVEMENTSUBGRADEANDNONDESTRUCTIVETESTING 55

loading. Figure 2b presents a drawing of the cylindrical specimen, 150-mm in

Downloaded from ascelibrary.org by UNIVERSITY OF FLORIDA on 11/10/15. Copyright ASCE. For personal use only; all rights reserved.

diameter by 150-mm high (approximately 6-in. in diameter by 6-in. high) under the
independently applied vertical and radial stresses and the instrumentation
consisting of LVDTs measuring axial and radial specimen deformations.

Figure 2. The University of Illinois FastCell (UI-FastCell) Triaxial Testing Device

Preliminary results obtained from four aggregates tested using the UI-
FastCell indicated definite directional dependency (anisotropy) of aggregate
stiffnesses (Tutumluer and Seyhan, 1999b). Tests were conducted on four
aggregates by pulsing only in vertical or horizontal direction the axial or radial
stress, respectively. The computed resilient moduli from these compression and
extension type loadings varied pronouncedly with the applied stress states. On the
other hand, when the same testing procedure was applied for testing a synthetic
specimen having isotropic material properties, the axial and radial moduli
computed from the measured resilient deformations remained unchanged. For
three of the materials having average particle sizes greater than the 4.76-mm (No. 4
sieve), higher vertical resilient moduli were attained at all 15 AASHTO T294-94
stress states, when compared to the horizontal resilient moduli. However, in the
case of a sandy gravel having an average particle size of 0.6-ram (0.02-in.) and a
high fines content (minus 0.075 mm or No. 200 sieve) of 13.9%, the computed
horizontal moduli were typically higher than the vertical moduli.

In a follow-up study, Tutumluer and Seyhan (1999a) conducted using the

UI-FastCell a total of 6 constant stress path compression and extension tests, which
indicated significant differences in the aggregate responses (see Figure 3a). A

Pavement Subgrade, Unbound Materials, and Nondestructive Testing

 56 PAVEMENTSUBGRADEAND NONDESTRUCTIVETESTING

well-graded crushed dolomite classified as A-l-a according to AASHTO soil

Downloaded from ascelibrary.org by UNIVERSITY OF FLORIDA on 11/10/15. Copyright ASCE. For personal use only; all rights reserved.

classification was selected for modulus testing and for studying the effects of stress
path loading on resilient moduli. The isotropic resilient moduli obtained from the
compression loadings were typically higher than the moduli computed from the
extension states as shown in Figure 3b. In the case that such critical loading
conditions exist in the field, lower granular layer stiffnesses would result in higher
responses associated with the reduced pavement performance.

(a) 15 AASHTO T294-94 stress states applied in 6 stress paths

(b) Variation of moduli from compression and extension stress paths with bulk stress

Figure 3. Stress Paths and Summary Results of UI-FastCell Testing Program

Conducted by Tutumluer and Seyhan (1999a)

Pavement Subgrade, Unbound Materials, and Nondestructive Testing

 PAVEMENT SUBGRADEAND NONDESTRUCTIVETESTING 57

Materials Tested In This Study

Downloaded from ascelibrary.org by UNIVERSITY OF FLORIDA on 11/10/15. Copyright ASCE. For personal use only; all rights reserved.

Aggregate Types and Properties

Twelve aggregate types with varying properties were selected for resilient
modulus (Mk) testing using the UI-FastCell. In the selection process, consideration
was given to both "good" and "poor" performing granular base/subbase materials
obtained from seven different states. These were the aggregates also studied in the
NCHRP 4-23 project, Performance Related Tests of Aggregates for Use in
Unbound Pavement Layers.' A realistic range of aggregate qualities and pro-
perties, such as average and top sizes, gradations (both uniform and well-graded
samples), particle shapes (rounded gravel to angular crushed stone), and fines
contents, i.e., materials less than 0.075 mm (No. 200 sieve) size, were represented.
The variations among the aggregate types and properties were considered essential
for studying the effects of material properties on the stress-induced anisotropic
behavior under the application of vertical and radial pulse loadings.

Table 1 reports the gradations and the material properties of the twelve
aggregates. The static strength properties were obtained from standard triaxial tests
(ASTM D 2850) conducted at confining pressures of 35, 69, and 104 kPa (5, 10, 15
psi) on samples compacted at optimum moisture states. Except for the PA Good
Quality, the materials studied can generally be considered as well-graded. The top
sizes vary from 25-mm (1-in.) to 51-mm (2-in.). Both crushed aggregates
composed of angular particles having rough surfaces such as the MN Fountain
Quarry, and rounded gravels such as the MN Shiely Elk River material can be
found in the material mix. There is a considerable variation in the fines content
values ranging from less than 1% in the case of PA Good Quality to 17.9% for the
TX Subgrade 1426 material.

Table 2 presents the maximum dry densities and the optimum moisture
contents as obtained from the modified Proctor tests (AASHTO T-180). The
achieved dry densities at optimum moisture contents, also shown in Table 2 for a
total of 24 samples, match closely with the maximum Proctor densities. The
achieved water contents are in general very close to the optimum values except for
those of the PA Good Quality material that had only 5% minus 4.75 mm (No. 4
sieve) size and 0.9% minus 0.075 mm (No. 200 sieve) size.

Sample Preparation

Cylindrical specimens, 150 mm in diameter by 150 mm high

(approximately 6-in. in diameter by 6-in. high), were prepared to fit in the
confinement chamber of the UI-FastCell for the repeated load triaxial testing. A
total of 24 specimens, two samples for each aggregate were prepared at the
optimum moisture content using a split aluminum compaction mold specifically
manufactured for use with the UI-FastCell. A nitrile membrane, 0.6 mm (0.025 in.)

Pavement Subgrade, Unbound Materials, and Nondestructive Testing

 58 PAVEMENT SUBGRADE AND NONDESTRUCTIVE TESTING
Downloaded from ascelibrary.org by UNIVERSITY OF FLORIDA on 11/10/15. Copyright ASCE. For personal use only; all rights reserved.

o~oo ~oo~o

~ oooooooooo~o
~ ~ ~
~ 1 7 ~6 1 7~6 1 7 6 1 7 6.~1 7 6 1 7 6 1 7 6 1 7 6 1 7

0000000000~0

~ ~ o o o o o o o o o o o

~ 0 0 0 0 0 0 0 ~ 0 0 0

~0
<
_~ , ~ , ~ , ~
~r~

e~
o

~ 0 ~

o ~ ~ 0 ~

zZ

Pavement Subgrade, Unbound Materials, and Nondestructive Testing

 PAVEMENT SUBGRADE AND NONDESTRUCTIVE TESTING 59
Downloaded from ascelibrary.org by UNIVERSITY OF FLORIDA on 11/10/15. Copyright ASCE. For personal use only; all rights reserved.

,-, %
r e'q ~ 0 0 ~ {'4 ~ cq e'4 oo ~q

< I II

,s

e~

II

d3

O
U [-.
o
0[-..

<
0 3 <
t

04
,.o
c~
F.

0
9 zz
*

Pavement Subgrade, Unbound Materials, and Nondestructive Testing

 60 PAVEMENTSUBGRADEAND NONDESTRUCTIVETESTING

thick, was attached to the bottom platen with an o-ring and the platen was placed in
Downloaded from ascelibrary.org by UNIVERSITY OF FLORIDA on 11/10/15. Copyright ASCE. For personal use only; all rights reserved.

the split mold. The split mold was assembled and the membrane was folded over
the top of the mold and secured with an o-ring. A vacuum line was attached to the
mold to hold the membrane tight against the mold. A non-woven geofabric was
placed on top of the bottom platen to prevent the drainage port from being clogged.
The aggregate mixed with required amount of water was placed in the mold in three
lifts and each lift was rodded 25 times using a standard rod for concrete testing.
The surface of each lift was leveled after rodding, and a full-faced compaction foot
with an air bubble leveling device on top was placed over the specimen.

A hand-held pneumatic vibratory compactor was used for compacting

aggregate specimens. Density was calculated by measuring the weight of material,
and the compacted thickness of each lift, referenced to the top of the mold. Each
lift was then scarified up to a depth of approximately 12-mm (0.5-in.), and the next
lift then placed, and compacted. After compaction, the final height and density of
specimen were noted.

After compaction, the vacuum was removed from the split mold and applied
to the bottom of the platen to create suction through the specimen thereby causing a
confinement by the membrane. The loading platen was placed at the top of
specimen. The split mold was then removed and a 0.3-mm (0.012-in.) thick nitrile
membrane was placed on the specimen and secured to the top and bottom platens
with o-rings. The second membrane was required because the first membrane
generally was punctured while compacting the specimen. Some powder was
applied on the second membrane to reduce the friction between the rubber
membrane of confining cell and the nitrile membrane of specimen while lowering
down, and raising up the confining cell. Next, the specimen was placed on the base
plate, centered by the pivot screw on the base plate, and fitted the hole at the
bottom of the bottom platen of specimen. Before lowering down the confining cell,
the loading plate was lowered down to make sure that the specimen was in upright
position by another pivot screw at the bottom of the loading plate, which fitted the
hole on top of the top platen. A 2-kPa (0.3-psi) hydrostatic seating pressure was
applied, and the vacuum was removed. The drainage port was left open to perform
the test under drained condition.

Resilient Modulus Testing

The cylindrical specimens were compacted by the pneumatic vibratory

compactor during the sample preparation stage. This compaction effort on the
specimens was assumed to represent the initial conditions of the granular layers in
the field just after pavement construction. Therefore, the specimens were not
conditioned before the actual testing sequence. Following the standard AASHTO
T294-94 procedure, each specimen was subjected to two sets of 15 triaxial stress
states that are typically less than the failure stress states. A haversine load
waveform was applied with a load pulse duration of 0.1-second (10-Hz), and a rest
period of 0.9-second. After the 2-kPa (0.3-psi) hydrostatic seating stress was

Pavement Subgrade, Unbound Materials, and Nondestructive Testing

 PAVEMENT SUBGRADEAND NONDESTRUCTIVETESTING 61

applied on the specimen, resilient modulus testing was conducted by pulsing first in
Downloaded from ascelibrary.org by UNIVERSITY OF FLORIDA on 11/10/15. Copyright ASCE. For personal use only; all rights reserved.

the vertical direction (direction 1) and then pulsing in the radial direction (direction
3) with the applied (pulsed) deviator stresses ~nd (n = 1 or 3). A second replicate
specimen was also tested for each material by changing the order of pulsing -
horizontal first and then vertical - to study the effects of stress history built in the
samples.

The pulsed deviator stresses, 6nd, ranged from 21 to 276 kPa (3 to 40 psi) in
both axial and radial directions whereas the hydrostatic pressures ranged from 21 to
138 kPa (3 to 20 psi). The applied stress ratios, total stress in any direction to
hydrostatic stress [(~nd + O'hydrostatic)/Ohydrostatic],ranged from 1.66 to 4. One hundred
load repetitions were applied at each stress state. Typically, the same vertical and
radial recoverable deformations were measured between the 50 th and 100 th load
repetitions.

Interpretation of Test Results

Resilient Moduli from UI-FastCell Triaxial Testing

Two samples of each aggregate were tested at near optimum moisture levels
under the application of directional dynamic loads. Each sample was subjected to a
total of 30 stress states with 15 deviator stresses applied in the vertical direction
and the other 15 in the horizontal direction. The vertical pulsing portion of the tests
was similar to the standard repeated load test (in accordance with AASHTO T-294-
94 and SHRP Protocol TP-46) with one-dimensional (l-D) dynamic loading
applied to the specimen in the axial direction. The horizontal pulsing portion of the
tests applied dynamic loading in the radial direction only to account for the resilient
response under extension states. The two applied stress states, vertical pulsing only
and horizontal pulsing only, are shown as the CCP1 (compression) and VCP5
(extension) tests in Figure 3a where C,d or Cr3d is the only pulsed axial or radial
stress, respectively.

The radial pulsed (deviator) stress, o3a, is not actually a I-D stress
application but rather all-around radial pressure acting on the specimen. The lateral
specimen surface that C3d is applied on has an area that is four times greater than
that of the circular top surface for the equal 150-mm (approximately 6-in.) diameter
and height of the cylindrical specimen. In other words, when applying a O3d similar
in magnitude to O~d, a dynamic horizontal load that is four times higher than that of
the previously cycled vertical load is actually applied on the specimen with each
repeated load. The order of pulsing direction, therefore, becomes especially
important for studying the stress history effects on the specimen.

The resilient modulus, MR, and the Poisson's ratio, v, are obtained from the
measured recoverable strains using axisymmetric stress-strain relations as follows:

Pavement Subgrade, Unbound Materials, and Nondestructive Testing

 62 PAVEMENTSUBGRADEAND NONDESTRUCTIVETESTING

1
(1)
Downloaded from ascelibrary.org by UNIVERSITY OF FLORIDA on 11/10/15. Copyright ASCE. For personal use only; all rights reserved.

E1 = [O'ld - 2VO3d ]
MR

E 3 = - -I [~3d(I-v)-v,~l,j] (2)
MR
where Old and O3d are the pulsed (dynamic) stresses only, and el and E3 are the
recoverable strains in the axial and radial directions, respectively. In using the
above equations to obtain resilient parameters as constants, the assumption is made
that for any individual stress state, the material behaves linearly and elastically.
The two equations must be solved simultaneously for the isotropic aggregate
properties, MR and v.

Vertical Pulsing Only: For the compression stress states with vertical
pulsing applied only (O3d = 0), the above stress-strain relations are simplified as,

1
E1 = " ~ R (~ld (3)

1
~3 = ~ R [- VGId] (4)

where M~ is the resilient modulus obtained from vertical pulsing only (o3a = O)
under various hydrostatic stresses and will be referred to as 'Vertical Resilient
Modulus' in this paper.

Horizontal Pulsing Only: For the extension stress states with horizontal
pulsing applied only (Old = 0), the simultaneous equations are simplified as,

El= ~-~[ 2VO3d] (5)

1
E3 = ~ R [O'3d(1- V)] (6)

where MhR is the resilient modulus obtained from horizontal pulsing only (trla -- O)
under various hydrostatic stresses and will be referred to as the 'Horizontal
Resilient Modulus' in this paper. Unlike in the vertical pulsing tests, the modulus
cannot be obtained independent of Poisson's ratio. Note that for an isotropic

Pavement Subgrade, Unbound Materials, and Nondestructive Testing

 PAVEMENT SUBGRADEAND NONDESTRUCTIVETESTING 63

material, there is only one resilient modulus regardless of the loading direction, M R
Downloaded from ascelibrary.org by UNIVERSITY OF FLORIDA on 11/10/15. Copyright ASCE. For personal use only; all rights reserved.

= M~ = M R , and one Poisson's ratio v.

Characterization of Aggregate Anisotropic Behavior

Unbound aggregate bases/subbases exhibit a directional dependency of

material stiffnesses under the stress states to which material is subjected
(Tutumluer, 1995; Tutumluer and Thompson, 1997a-b). This directional
dependency of resilient granular material behavior can be properly represented as
power functions of the stress states in the granular layer (Hicks and Monismith,
1971; Uzan, 1985). Due to their simplicity and good prediction ability, the two
resilient modulus models were recommended in the past for use as practical models
to characterize granular bases in flexible pavements (Tutumluer, 1995; Tutumluer
and Thompson 1997a-b). These models, K-0 and Uzan, also used in this study are
as follows:

K-O Model (Hicks and Monismith, 1971)." MR= K (7)

Uzan Model (Uzan, 1985): MR-K,/O/2//K3 (8)

tPo) tPo)
where 0 is bulk stress, po is the unit reference pressure (1 kPa or 1 psi), and K, Kl,
K2, K3, and n are material constants obtained from regression analyses performed
on UI-FastCell repeated load triaxial test results.

Bulk stress, 0, refers to the total pressure applied on the specimen including
both the hydrostatic pressure, ~hydrostatic, and the deviator stress, ~nd (n = 1 or 3).
Depending upon the pulsing direction, the bulk stress 0 was computed either from
the equation, 0 = Old + 3~hydrostatic, for the vertical pulsing (compression) tests or e
= 2~3cl + 3~hydrostatic for the horizontal pulsing (extension) tests. In what follows,
the vertical modulus is then computed using the dynamic deviator stress cy~ and the
horizontal modulus is computed using the dynamic deviator stress ~3d. Therefore,
either the vertical or the horizontal modulus that is characterized using Equations 7
and 8 is always predicted in the direction of the pulsating load.

The experimental resilient moduli and Poisson's ratio were calculated from
the measured axial and radial recoverable strains using Equations 3 to 6. The
material constants employed in the stress dependent models were then determined
from regression analyses of the experimental resilient moduli. The K-0 and Uzan
model parameters and the correlation coefficients (R2s) obtained for only three of
the materials, IN DOT Section #2421, PA DOT Good Quality, and TX DOT

Pavement Subgrade, Unbound Materials, and Nondestructive Testing

 64 PAVEMENT SUBGRADE AND NONDESTRUCTIVE TESTING

Section #13 Subgrade 1426, are presented in Table 3 for the two samples tested of
Downloaded from ascelibrary.org by UNIVERSITY OF FLORIDA on 11/10/15. Copyright ASCE. For personal use only; all rights reserved.

each material.

Table 3. Stress Dependent Model Parameters for Horizontal and Vertical Moduli

TX IN PA
Material Type Subgrade #1426 Section #2421 Good Quality
Sample Number Tested* 1 2 1 2 1 2

K
K-0 (MPa) 339.00 349.51 546.63 514.58 341.73 330.28

Model n 0.72 0.78 0.82 0.64 0.83 0.61

Vertical R2 0.956 0.960 0.977 0.972 0.956 0.972
Modulus KI
v (MPa) 234.65 250.40 433.98 406.17 282.67 257,98
MR Uzan
I~ 0,99 1.02 0.99 0.81 0.97 0.80
Model
K3 -0.30 -0.27 -0.19 -0.19 -0.15 -0.20
R2 0,999 0.990 0.990 0.994 0.964 0.998

K
K-0 (MPa) 345.43 326.88 210.39 168.98 151.08 252.80

Model n 0.69 0.64 0.95 0.84 0.94 1.22

Horizontal Rz 0.875 0.807 0.930 0.923 0.865 0.849
Modulus
Kl
Uzan (MPa) 154.73 111.20 82.19 72.63 40.18 46.41

Model K2 1,24 1.38 1.59 1.43 1.85 2.38

K3 -0.56 -0.75 -0.65 -0.59 -0.92 -1.18
R2 0.966 0.979 0.997 0.993 0.995 0.975
* Sample I : First compression then extension
Sample 2: First extension then compression

Figures 4-6 show for the three aggregates the variations of the vertical and
horizontal resilient moduli with increasing bulk stresses. The data points represent
the experimental moduli as calculated from Equations 3 to 6 for various stress
states. Overall, the anisotropic resilient moduli for each granular material increased
as the applied hydrostatic stresses increased. The K-0 and Uzan model predicted
moduli are also shown as trend lines in the figures. The K-0 model, shown with
dashed lines, performed a more general curve fitting to the experimental moduli.

Pavement Subgrade, Unbound Materials, and Nondestructive Testing

 PAVEMENT SUBGRADE AND NONDESTRUCTIVE TESTING 65
Downloaded from ascelibrary.org by UNIVERSITY OF FLORIDA on 11/10/15. Copyright ASCE. For personal use only; all rights reserved.

[..

O
[..

r
O
t~
9

t"q

oo

~ Z

,4

Whereas, the Uzan model predictions, shown with solid lines in the figures,
followed more closely the experimental values. This is due to the use of two
independent stress states, (~hydrostatic and o'n~, in the Uzan models that resulted in

Pavement Subgrade, Unbound Materials, and Nondestructive Testing

 66 PAVEMENTSUBGRADEAND NONDESTRUCTIVETESTING

better statistical fit with higher regression coefficients (R2) (see Table 3). Hence,
Downloaded from ascelibrary.org by UNIVERSITY OF FLORIDA on 11/10/15. Copyright ASCE. For personal use only; all rights reserved.

the Uzan type models properly modeled the effect of varying deviator stress at one
hydrostatic pressure on the vertical and horizontal resilient moduli as indicated by
the individual best-fit trend lines for the five different hydrostatic pressure levels.

~5
O

[-.,
Q

9. ~ .-,
e.- r
t~ ,.-I

".~ 0

;>

Pavement Subgrade, Unbound Materials, and Nondestructive Testing

 PAVEMENT SUBGRADE AND NONDESTRUCTIVE TESTING 67
Downloaded from ascelibrary.org by UNIVERSITY OF FLORIDA on 11/10/15. Copyright ASCE. For personal use only; all rights reserved.

<D

~D
[...

~5
O

b-,
E
O

,.-I

e-.

-r" "~
O

,.Q

i.

oO
0
"~, X

,d

On the contrary to the isotropic material property assumption, the vertical

and horizontal moduli obtained from compression and extension type loadings
were not similar for the 12 materials tested. For all samples tested of the "good

Pavement Subgrade, Unbound Materials, and Nondestructive Testing

 68 PAVEMENTSUBGRADEAND NONDESTRUCTIVETESTING

quality" materials, PA DOT and IN DOT Section #2421 shown in Figures 4 and 5,
Downloaded from ascelibrary.org by UNIVERSITY OF FLORIDA on 11/10/15. Copyright ASCE. For personal use only; all rights reserved.

higher vertical resilient moduli were consistently attained at all 15 AASHTO T294-
94 stress states, when compared to the horizontal resilient moduli. The gradation
properties influence these anisotropic responses. These supposedly good
performing base/subbase materials usually have low to moderate amounts of fines
and/or their average particle sizes, Ds0 corresponding to 50% passing, are greater
than the 4.75 mm (No. 4 sieve) size. On the other hand, materials having high
fines, such as the TX Subgrade 1426 (see Figure 6), and/or average particle sizes,
Dso, greater than the 4.75 mm (No. 4 sieve) size, such as the MN Elk River, had
higher horizontal moduli than the vertical ones.

The effects of stress history, i.e., order of pulsing direction or application of

compression/extension state first, on the samples are also indicated by the two plots
given for each aggregate in Figures 4 to 6. A direct comparison can be made
between the computed moduli at various stress states by simply extending the
modulus values horizontally for samples 1 and 2. The most significant and
consistent stress history effect was observed for the "good quality" materials
having moderate amount of fines, especially for IN Section #2421 as indicated in
Figure 4. This material had much higher vertical resilient moduli computed from
the second replicate tests (sample 2 with extension states applied first) when
compared to the values computed from sample I. This was due to strengthening of
the material significantly in the vertical direction under the higher horizontal loads
applied on the specimens at similar stress states. Previous studies also reported
such findings that linked the important effects of stress history to the presence of
moderate to high fines in the samples (Tutumluer and Seyhan, 1999b).

Note that the uniformly graded PA Good Quality material also exhibited a
similar behavior in which higher vertical resilient moduli were computed from the
second sample and yet, this material has only 0.9% fines, Presumably, this material
experienced more of a particle reorientation (sliding, translation, etc.) under higher
horizontal loads than a direct effect of the stress history, i.e., order of pulsing
direction affecting the behavior. A closer look at the results shown in Figure 5
reveals that not only the vertical but also the horizontal moduli increased
considerably going from sample 1 to sample 2. This is a good indication that a
reorientation of the particles probably took place to result in better aggregate
interlock. As a result, higher stiffness properties were attained for transferring the
high magnitudes of horizontally applied loads. No significant changes were
observed in the anisotropic moduli when the order of pulsing direction was
changed in the replicate tests for the "poor quality" materials having high fines
such as the TX Section #13 Subgrade #1426.

Figure 7 shows varying degrees of stiffness anisotropy, i.e., horizontal to

vertical modular ratios, plotted with the applied deviator stresses for the "first
compression then extension" type tests on the 12 aggregates. For ease in
visualization and establishing trends in behavior, the modular ratios shown were

Pavement Subgrade, Unbound Materials, and Nondestructive Testing

 PAVEMENTSUBGRADEAND NONDESTRUCTIVETESTING 69

obtained only at three stress states referred to here as low, intermediate, and high.
Downloaded from ascelibrary.org by UNIVERSITY OF FLORIDA on 11/10/15. Copyright ASCE. For personal use only; all rights reserved.

Increasing or decreasing trends with increasing deviator stress are clearly observed
for different materials. The "poor quality" materials are indicated in Figure 7 by
the high modular ratios shown in the upper portion of the graph. In addition,
modular ratios of all the "poor quality" materials having relatively high fines
content, consistently decrease as the stresses applied on the specimen increase from
low to high states. Furthermore, those materials that show increasing modular
ratios with increasing stress levels are the "good quality" materials. In general, the
obtained results are in very good agreement with the previous findings that
indicated horizontal to vertical modular ratios ranging from 0.9 to 4 for cohesive
soils such as clays and ratios as low as 0.2 for sands and aggregates (Gazetas, 1982;
Graham and Houlsy, 1983; Lo and Lee, 1990; Jiang et al., 1997, Tutumluer and
Seyhan, 1999b).

Summary/Conclusions
An innovative triaxial testing machine, referred to as University of Illinois
FastCell (UI-FastCell), was used for determining the directional dependency
(anisotropy) of resilient properties of 12 "good" and "poor" performing
base/subbase materials obtained from seven states. This could be achieved by
independently pulsing deviator stresses in the vertical or horizontal direction to
apply compression and extension type loadings on the specimen and measuring
both the vertical and radial specimen deformations.

Nonlinear stress dependent models, K-0 and Uzan, were adequately

developed from measured axial and radial strains for characterizing the anisotropic
aggregate stiffnesses under compression and extension type loadings. The Uzan
model, having the additional deviator stress term in the formulation, followed more
closely the experimental moduli.

For all samples tested of the "good quality" materials, higher moduli were
consistently attained under compression type loadings at all 15 AASHTO T294-94
stress states, when compared to the ones obtained under extension states. On the
other hand, "poor quality" materials having high fines and/or average particle sizes,
Ds0, greater than 4.75 mm (No. 4 sieve) generally had higher moduli under
extension states. The most significant stress history effect was observed for the
"good quality" materials having moderate amount of fines.

Much lower anisotropic (horizontal to vertical) modular ratios were

obtained for the "good quality" materials than those of the "poor quality" ones at
low, intermediate, and high stress states. As the stress states applied on the
specimen increased from low to high states, the modular ratios consistently
increased for the "good quality" materials and consistently decreased for the "poor
quality" materials.

Pavement Subgrade, Unbound Materials, and Nondestructive Testing

 70 PAVEMENTSUBGRADEAND NONDESTRUCTIVETESTING
Downloaded from ascelibrary.org by UNIVERSITY OF FLORIDA on 11/10/15. Copyright ASCE. For personal use only; all rights reserved.

Figure 7. Variations of Anisotropic Modular Ratios with Applied Deviator Stresses

The amount of fines in an aggregate makes a significant impact on the

anisotropic behavior. The good performing base/subbase material should harden
and get stronger under applied stresses increasing its horizontal stiffness relative to
the vertical stiffness thus reducing its tendency spread laterally under wheel loads.
To account for the most damaging field stress states on a pavement element, it is

Pavement Subgrade, Unbound Materials, and Nondestructive Testing

 PAVEMENTSUBGRADEAND NONDESTRUCTIVETESTING 71

necessary to include in our laboratory testing and material characterization the

Downloaded from ascelibrary.org by UNIVERSITY OF FLORIDA on 11/10/15. Copyright ASCE. For personal use only; all rights reserved.

extension type loading conditions. The extension states have been shown to
generally give much lower resilient moduli than the compression states, which
would result in higher critical pavement responses associated with reduced
pavement performance.

Acknowledgments

The authors would like to thank Drs. Jim Hall and Athar Saeed of the
Applied Research Associates, Inc., Vicksburg, MS for their support of the
laboratory-testing program and for providing the aggregate samples also used in the
NCHRP 4-23 project, 'Performance Related Tests of Aggregates for Use in
Unbound Pavement Layers.'

References

AASHTO T294-94 (1995). Resilient Modulus of Unbound Granular Base/Subbase

Materials and Subgrade Soils - SHRP Protocol P46. Standard Specifications
for Transportation Materials and Methods of Sampling and Testing,
Seventeenth Edition, American Association of State Highway and
Transportation Officials, Washington D.C.
Garg, N., Tutumluer, E., and Thompson, M.R. (1998). Structural Modeling
Concepts for the Design of Airport Pavements for Heavy Aircraft. In
Proceedings of the 5th International Conference on the Bearing Capacity of
Roads and Airfields, Trondheim, Norway, July, pp. 115-124.
Gazetas, G. (1982). Stresses and Displacements in Cross-Anisotropic Soils. Journal
of Geotechnical Engineering, ASCE, Vol. 108, No. GT4, April, pp. 532-553.
Graham, J. and G.T. Houlsby (1983). Anisotropic Elasticity of A Natural Clay.
Geotechnique, Vol. 33, No. 2, pp. 165-180.
Hicks, R.G. and C.L. Monismith. (1971). Factors Influencing the Resilient
Properties of Granular Material. In Transportation Research Record 345, TRB,
National Research Council, Washington D.C., pp. 15-31.
Jiang, G.-L., Tatsuoka, F., Flora, A., and Koseki, J. (1997). Inherent and Stress-
State-Induced Anisotropy in Very Small Strain Stiffness of A Sandy Gravel.
Geotechnique, Vol. 47, No. 3, pp. 509-521.
Lo, S-C.R. and I.K. Lee (1990). Response of Granular Soil along Constant Stress
Increment Ratio Path. Journal of Geotechnical Engineering, ASCE, Vol. 116,
No. 3, pp. 355-376.
Tutumluer, E. (1995). Predicting Behavior of Flexible Pavements with Granular
Bases. Ph.D. Dissertation, School of Civil and Environmental Engineering,
Georgia Institute of Technology, Atlanta, GA, September.
Tutumluer, E. and M.R. Thompson (1997a). Anisotropic Modeling of Granular
Bases in Flexible Pavements. In Transportation Research Record 1577, TRB,
National Research Council, Washington D.C., pp. 18-26.

Pavement Subgrade, Unbound Materials, and Nondestructive Testing

 72 PAVEMENT SUBGRADEAND NONDESTRUCTIVETESTING

Tutumluer, E. and M.R. Thompson (1997b). Granular Base Moduli for Mechanistic
Downloaded from ascelibrary.org by UNIVERSITY OF FLORIDA on 11/10/15. Copyright ASCE. For personal use only; all rights reserved.

Pavement Design. Proceedings of the ASCE Airfield Pavement Conference,

Seattle, Washington, August 17-20, 1997, pp. 33-47.
Tutumluer, E. and U. Seyhan (1999a). Stress Path Loading Effects on Granular
Material Resilient Response. Invited Paper, In Proceedings of the European
Community COST 337 Workshop on Modeling and Advanced Testing for
Unbound Granular Materials, Instituto Superior T6cnico - IST, Lisbon,
Portugal, January 21-22, 1999.
Tutumluer, E. and U. Seyhan (1999b). Laboratory Determination of Anisotropic
Aggregate Resilient Moduli Using A New Innovative Test Device. In
Transportation Research Record 1687, Journal of the Transportation Research
Board, National Research Council, Washington, D.C., pp. 13-21.
Uzan, J. (t985). Characterization of Granular Materials. In Transportation
Research Record 1022, TRB, National Research Council, Washington D.C.,
1985, pp. 25-59.

Pavement Subgrade, Unbound Materials, and Nondestructive Testing