Resilient behavior of compacted subgrade soils under

the repeated triaxial test

Daehyeon Kim
a,
*
, Jong Ryeol Kim
b,1
a
Indiana Department of Transportation (INDOT), Research and Development, 1205 Montgomery Street, West Lafayette, IN 47906, USA
b
Department of Civil Engineering, Chonnam National University 300 Yongbong-Dong, Buk-Gu, Kwangju 500-757, South Korea
Received 21 March 2006; accepted 10 July 2006
Available online 14 September 2006
Abstract
Subgrade soils are very important materials to support highways. Resilient modulus (M
r
) has been used for characterizing stress-
strain behavior of subgrades subjected to repeated trac loadings. Recently the repeated triaxial test procedure has been upgraded
through AASHTO T 307. Since the testing procedure is still complex, the testing has not been widely implemented in practice. In order
to evaluate resilient behavior of compacted subgrades soils, the repeated triaxial test and the unconned compressive test as well as some
fundamental property tests were conducted. In this study, the applicability of a simplied procedure with a conning pressure of 13.8 kPa
and deviator stresses of 13.8, 27.6, 41.4, 55.2, 69 and 103.4 kPa was investigated on the typical sandysiltyclay and siltyclay subgrade
soils encountered in Indiana. The results obtained from the simplied procedure are comparable with those obtained from AASHTO T
307 which calls for 15 combinations of stresses. This shows the simplied procedure to be feasible and eective for design purpose. Some
soils compacted wet of optimum moisture content showed an excessive permanent deformation. This phenomenon was investigated by
the comparison of the unconned compressive test and the repeated triaxial test results. For soils exhibiting excessive permanent defor-
mation, use of deformed length is desirable for more accurate calculation of M
r
. Usually the soils compacted dry of optimum shows the
largest M
r
for sandysiltyclay soils due to capillary suction, but it is not necessarily true for siltyclay soils. A predictive model to esti-
mate regression coecients k
1
, k
2
, and k
3
using 11 soil variables obtained from the soil property tests and the standard Proctor compac-
tion tests was developed. The predicted regression coecients compare well with measured ones.
2006 Elsevier Ltd. All rights reserved.
Keywords: Resilient modulus; Siltyclay soils; Predictive model; Soil properties; Regression coecients; Permanent deformation
1. Introduction
1.1. Background and research objective
Since the AASHTO Guide for Design of Pavement
Structures in 1986 recommended highway agencies to
use a resilient modulus (M
r
) value obtained from a
repeated triaxial test for the design of pavements, numer-
ous eorts have been made in obtaining more accurate,
straightforward, and appropriate M
r
values which are rep-
resentative of eld conditions. In the past or even in the
present, most highway agencies have used the California
Bearing Ratio (CBR) to characterize subgrades in the
design of pavements and to correlate it with the resilient
modulus. CBR, however, is a static property that cannot
account for the actual response of the subgrade under the
dynamic loads of moving vehicles [1].
Although the AASHTO design guide requires designers
to use one resilient modulus value representative for a
given subgrade considering seasonal variation, it is not easy
to obtain the resilient modulus by performing a standard
repeated triaxial test due to its complex, time-consuming
and costly testing procedure [24]. Due to its complexity
0950-0618/$ - see front matter 2006 Elsevier Ltd. All rights reserved.
doi:10.1016/j.conbuildmat.2006.07.006
*
Corresponding author. Tel.: +1 765 463 1521; fax: +1 765 497 1665.
E-mail addresses: dkim@indot.state.in.us (D. Kim), jrkim@chonna-
m.ac.kr (J.R. Kim).
1
Tel.: +82 62 530 1654; fax: +82 62 530 1650.
www.elsevier.com/locate/conbuildmat
Construction and Building Materials 21 (2007) 14701479
Construction
and Building
MATERIALS
and diculty, many correlations have been made between
M
r
values from the repeated triaxial test and measurements
obtained from nondestructive eld testing methods, such as
the Cone Penetration Test (CPT), the Dynamic Cone Pen-
etration Test (DCPT), the Falling Weight Deectometer
(FWD), and the Plate Load Test (Plate Load Test). Of
these methods, CPT is one of the most frequently used
methods due to its economy, reliability and repeatability
[1]. Also, at small strain levels (i.e. less than 0.1%), some
laboratory tests such as unconned compression test [5
7] and static triaxial test [8] were suggested to nd some
alternatives to the repeated triaxial test. A simplied testing
procedure was also suggested by decreasing the number of
conning and deviator stresses [2].
It is noted that the AASHTO Design guide recommends
highway agencies to use representative conning and devi-
ator stresses in subgrade layers under trac loading condi-
tions. Therefore, it is necessary to investigate the range of
conning and deviator stresses resulting from the trac
loadings and to account for such reasonable stress levels
in the repeated triaxial test. Over- or underestimation of
the stress levels in the subgrades will lead to erroneous
resilient modulus results [9]. Through a repeated triaxial
test, as one resilient modulus corresponding to the repre-
sentative conning and deviator stress is needed for a given
subgrade in designing a pavement, the complex testing pro-
cedure can be simplied for practical design purposes.
The current standard test method to determine the resil-
ient modulus is described in the AASHTO T 307-99 which
has recently been upgraded from AASHTO T 294-94 and
AASHTO T 274. Most literature is limited to AASHTO
T 294-94 and AASHTO T 274, so no literature on the eval-
uation of AASHTO T 307-99 appears to be available. In
AASHTO T 307-99, eld conditions are simulated by con-
ditioning and postconditioning (i.e., main testing). Condi-
tioning consists of 5001000 load applications at a
conning stress of 41.4 kPa and a deviator stress of
27.6 kPa. In addition, main testing is performed at three
levels of conning stresses (in the order of 13.8, 27.6 and
41.4 kPa) for which each ve levels of deviator stresses
(13.8, 27.6, 41.4, 55.2 and 69 kPa) are applied, resulting
in 15 steps of load applications. It classies soil types into
types 1 and 2 materials. Granular soils and cohesive soils
are categorized as types 1 and 2, respectively. This test
applies the same procedure for both granular and cohesive
subgrades and is done under a drained condition only.
There have been two research projects [10,7] on the resil-
ient modulus tests focusing on the evaluation of the resil-
ient behavior of typical subgrade soils in Indiana and its
implementation. In order to achieve more realistic design
of subgrade, the Indiana Department of Transportation
(INDOT) has required the resilient modulus to be used.
For most of new alignment and reconstruction of the road-
way, the repeated triaxial test results are required on com-
pacted samples at moisture contents (OMC and two
percent greater than OMC). However, the repeated triaxial
tests are being performed by only several specialized labo-
ratories due to its complex and dicult procedure. The
complexity and diculty of performing the repeated triax-
ial test prevented INDOT and others from performing the
repeated triaxial test as a routine test. For small projects,
the CBR test is performed and an empirical correlation is
used to convert CBR to M
r
.
In this study, an attempt to simplify the procedure of the
standard resilient modulus test (AASHTO 307-99 or
AASHTO 294-94) is made for typical compacted subgrade
soils. The objective of this study is to simplify the existing
repeated triaxial test procedure, to evaluate the resilient
behavior of Indiana subgrades and to develop a predictive
model for estimation of the resilient modulus, allowing
construction of a database on the resilient modulus and
helping engineers in the evaluation of M
r
.
1.2. Resilient modulus
Resilient modulus is an important material property of
subgrade soils and is an input parameter in the design of
pavements. It is based on the recoverable strains after a ser-
ies of combination of conning and deviator stresses in the
repeated triaxial test is applied to a soil specimen in order
to take into account nonlinear behavior of subgrade soils
under the trac loadings. Resilient modulus is dened as:
M
r

r
d
e
r
1
where M
r
is the resilient modulus, r
d
is the repeated devia-
tor stress, e
r
is the recoverable axial strain.
An idealized schematic for a conning stress and a devi-
ator stress is shown in Fig. 1 where e
a
is the irrecoverable
axial strain and e
r
is the recoverable axial strain. An under-
lying assumption in Fig. 1 is that so long as the deviator
stress applied is not in excess of the shear strength of the
soil, only recoverable strain occurs after a numerous num-
ber of repetitions. However, this assumption may not be
D
e
v
i
a
t
o
r

s
t
r
e
s
s

Strain (%)
Mr

r
a
Fig. 1. Denition of resilient modulus for a conning and a deviator
stress.
D. Kim, J.R. Kim / Construction and Building Materials 21 (2007) 14701479 1471
true for all soils subjected to the trac loading. For some
cohesive soils, excessive permanent strain can often be
observed while performing the repeated triaxial test even
under the typical trac loading condition (i.e., 80 kN (18
kips) single axle load). As indicated in Eq. (1), the larger
the recoverable deformation, the smaller the resilient mod-
ulus (i.e., weak subgrade). This indicates that if the recov-
erable deformation is quite large, even the continuous
elastic deformation can damage to the integrity of the high-
ways as a result of repeated loading.
In AASHTO T307, M
r
is calculated using the original
length of the specimen tested. For soils having small per-
manent deformation, this would be acceptable, but for soils
showing larger permanent deformation, the use of the
deformed length of the specimen and corrected area would
be helpful for more accurate M
r
to be obtained. Also, it is
intuitively noticed from Eq. (1) that typical range of the
recoverable strain level measured during the repeated triax-
ial test is on the order of 10
3
10
4
, which suggests how
accurately a measurement of displacement is required.
1.3. Resilient behavior of soils
In general, the resilient modulus of subgrades is aected
by the following factors: deviator stress; conning pressure;
water content and dry density; method of compaction;
thixotropy; and freeze-thaw cycles. The resilient modulus
of cohesive soils is mainly a function of the applied devia-
tor stress (i.e., little eect of conning pressure), when a
single conning stress level is considered [11]. On the other
hand, the resilient modulus of granular materials increases
with increasing conning pressure. For cohesive subgrades,
there is a general consensus that at low levels of repeated
deviator stress, the resilient modulus decreases signicantly
as the deviator stress increases, while at greater levels of
deviator stress, the resilient modulus either decreases
slightly or reaches constant values [6,12,13]. Generally,
there is a breakpoint resilient modulus corresponding to
the resilient modulus at a deviator stress of 41.4 kPa [12].
This breakpoint characterizes the behavior of these soils
under repeated loads and might be dierent depending on
the subgrades.
It appears that there is some disagreement between
researchers on conning pressure for cohesive subgrades.
It is noted that Lee et al. [14] obtained considerably dierent
values for dierent conning stresses, while others showed
little eect of conning stresses on the modulus of cohesive
soils [6]. It is also noted that Thompson and Robnett [12] in
their study used zero conning pressure with change in devi-
ator stress. As such, the eect of conning pressure on resil-
ient modulus of cohesive subgrades has not been
thoroughly studied based on very careful investigation of
drainage mechanism during testing and detailed soil types
depending on clay contents and moisture contents.
Most researchers have shown that there is little eect of
moisture content on resilient modulus of granular sub-
grades [6,15], while there is signicant eect of water con-
tent on resilient modulus of cohesive subgrades (i.e. the
higher the moisture content, the less the resilient modulus)
[6].
1.3.1. M
r
in the design
Although M
r
values are obtained from a repeated triax-
ial test, the nal design M
r
value for a certain subgrade is
determined through an additional step. It is generally
known that moisture and freeze-thaw cycles signicantly
aect the subgrade support capability and their eects
should be taken into account [3]. In the AASHTO design
guide, seasonal or monthly M
r
values are required to
account for the monthly variations of M
r
values. Elliot
and Thornton [3] suggested a method to account for the
monthly variations of M
r
values based on the relationships
between M
r
and moisture contents, and moisture contents
and months. But their methodology was not very sophisti-
cated as the eect of temperature was not taken into
account. According to the study done by Lee et al. [14],
the average M
r
value for four frozen Indiana soils was
obtained. Based on climatic data in Indiana, they suggested
a tentative representation of subgrade M
r
throughout the
year. The M
r
starts to increase in December and a high
M
r
is maintained until March. M
r
decreases abruptly in
April due to thawing, being followed by ve months of a
thaw recovery period. Once the seasonal variation in M
r
is estimated, the eective roadbed resilient modulus (i.e.,
Design M
r
) can be obtained following the procedure spec-
ied in AASHTO design guide.
2. Testing program
A total of eleven soils (four sandysiltyclay and seven
siltyclay) encountered in Indiana were used in the testing
program. The testing program consisted of the sieve analy-
sis, the Atterbergs limit test, the standard Proctor compac-
tion test, the unconned compressive test, and the repeated
triaxial test. Fig. 2 shows the particle size distribution and
Table 1 shows material properties of these soils. The soils
(I65-146, I65-158, I65-172 and Dsoil) are sandysiltyclay
soils and the rest of the soils in Table 1 are siltyclay soils.
For each soil, the standard Proctor compaction tests
were conducted. Three samples for the repeated triaxial
testing were made at dierent water contents which are
dry of optimum (95% relative compaction), optimum
(100% relative compaction), and wet of optimum (95% rel-
ative compaction). After the repeatability of the repeated
triaxial test and unconned test was ensured, the main test-
ing was performed. Throughout the paper, dry of opti-
mum, optimum and wet of optimum correspond to 95%
relative compaction (dry side), 100% compaction, 95% rel-
ative compaction (wet side), respectively. A wide range of
water content was used to account for the possible range
of lower and upper bounds of M
r
values. Note that the per-
cent relative compaction is dened as the percentage of the
dry unit weight (c
d
) divided by the maximum dry unit
weight (c
dmax
) in the compaction curve.
1472 D. Kim, J.R. Kim / Construction and Building Materials 21 (2007) 14701479
For preparation of a sample for a repeated triaxial test,
a compaction mold, specially constructed, with a diameter
of 7.1 cm and a height of 15.2 cm was used. Five layers of
compaction were done with the same compaction energy as
the standard Proctor compaction test. Compaction curves
for all of the soils tested are shown in Fig. 3. As can be
in Fig. 3, for sandysiltyclay soils the dry unit weight is
in the range of 18.520 kN/m
3
and the optimum water con-
tent ranges between 9% and 13%, while for siltyclay soils
the dry unit weight ranges from 15 to 18 kN/m
3
and the
optimum water content ranges between 12% and 23%.
An automated repeated triaxial test device made by
Geocomp. Corp. was used for the repeated triaxial test.
Air pressure was used for applying a conning pressure.
After a specimen is put into a triaxial chamber, the
repeated triaxial test is begun. The repeated triaxial test is
completed after a series of loading combinations as speci-
ed in AASHTO T307.
3. Discussion of test results
3.1. Proposed simplied procedure vs. AASHTO T307
As mentioned previously, the current AASHTO T307
calls for 15 steps of repeated loading. The primary reason
for that is to apply the trac loading in a wide range cov-
ering the typical loadings. In the design of pavements, resil-
ient modulus values of subgrades corresponding to the
S
a
n
d

L
i
m
i
t
S
i
l
t

L
i
m
i
t
C
l
a
y

L
i
m
i
t
0%
10%
20%
30%
40%
50%
60%
70%
80%
90%
100%
0.0001 0.001 0.01 0.1 1 10
Particle Diameter (mm)
P
e
r
c
e
n
t

F
i
n
e
r
I65-146 I65-158 I65-172 Dsoil #1 #2 #3 #4 SR 19 US 41 Bloomington
Fig. 2. Particle size distribution for soils used.
Table 1
Material properties for soils used
Soil Gravel (%) Sand (%) Silt (%) Clay (%) LL (%) PI (%) AASHTO USCS
I65-146 8 34 40 18 18.5 5.2 A-4 CLML
I65-158 8 38 44 10 18.2 4.6 A-4 CLML
I65-172 16 33 33 18 24.2 14.7 A-6 CLML
Dsoil 0 17 61 22 26 6.2 A-4 CLML
#1soil 0.3 4.8 52.3 42.6 50 23 A-7-6 CH
#2soil 2.6 20.5 52.7 24.2 39 16 A-6-12 CL
#3soil 8.7 20.6 62.6 8.1 40 15 A-6-10 CL
#4soil 2.5 23.2 59.8 14.5 43 21 A-7-6 CL
SR19 3.2 21.5 55.4 19.9 33 16 A-6-10 CL
US41 0.9 19.6 58.1 21.4 28 9 A-4-6 CL
Bloomington 0.3 3.2 60.6 35.9 46 26 A-7-6 CL
Note: LL, liquid limit; PI, plasticity index; AASHTO, American Association of State Highway Transportation Ocials; USCS, unied soil classication
system; CL, clay with low plasticity; CH, clay with high plasticity; CLML, silty clay.
D. Kim, J.R. Kim / Construction and Building Materials 21 (2007) 14701479 1473
representative stress levels on top of the subgrades are
important because these values should be used for design
parameters. Generally, the level of conning stress on top
of the subgrades induced by 80 kN (18 kips) Equivalent
Single Axle Load (ESAL) would be about 13.827.6 kPa
[3]. In our study, several multi-layered elastic analyses for
typical cross-sections using ELSYM5 showed the
13.8 kPa as a minimum conning pressure for typical Indi-
ana roads [16]. Therefore, one attempt was made to make
the procedure quicker and easier. As a consequence, it was
determined that a conning stress of 13.8 kPa and deviator
stresses of 13.8, 27.6, 41.4, 55.2 and 69 kPa were appropri-
ate for the simplied M
r
procedure. Note that the proposed
simplied procedure consists of ve steps while the current
procedure requires 15 steps.
Fig. 4a and b show the comparisons of the M
r
values
between the simplied and the AASHTO procedures,
where those soils were compacted at optimum moisture
contents for I65-158 and I65-172. It is clearly seen in
Fig. 4a and b that the higher the conning stress, the higher
the resilient modulus value, which is the typical behavior of
subgrade soils. In Fig. 4a the number of repetition in the
conditioning stage and the main testing was the same as
the one in the AASHTO T307, while in Fig. 4b the number
of repetitions both in the conditioning stage and the main
testing stage was reduced by half the number as per
AASHTO T307. The M
r
values obtained from the two
methods are in good agreement (R
2
= 0.9 for Fig. 4a and
R
2
= 0.76 for Fig. 4b). This means that the simplied pro-
cedure can be appropriately used for estimation of M
r
val-
ues in place of the current repeated triaxial test method,
AASHTO T 307.
3.2. M
r
values for dry, OMC and wet water contents
In general, the repeated triaxial test is performed at opti-
mum moisture content (OMC) or 2% of the OMC. In the
eld, however, compaction control is conducted by the
percent relative compaction with respect to the standard
Proctor compaction curve. Ninety-ve percent relative
compaction is usually incorporated for compaction control
of subgrades, which allow some cases where water contents
exist dry of optimum or wet of optimum. In order to
account for such eld conditions, the repeated triaxial test
was performed on soils compacted dry of optimum, opti-
mum and wet of optimum. It should be noted that the dif-
ference in water contents between them is considerably
large, approximately 512%, which is dependent on the
shape of the compaction curve.
It is very important to distinguish the meaning of sti-
ness and strength of the soil. The resilient modulus of sub-
grade soil does not represent its strength but stiness. For
instance, a soil having a higher strength than the other does
not necessarily show higher stiness; it may show either
higher or lower stiness. Table 2 shows the measured M
r
values for soils compacted dry of optimum, optimum and
wet of optimum at a conning stress of 13.8 kPa and a
deviator stress of 41.4 kPa. This indicates that subgrade
soil must be compacted adequately at near the optimum
moisture content in order for the subgrade to better sup-
port the highways.
3.3. Sandysiltyclay soils
As indicated in Table 2, for all of the four sandysilty
clay soils tested, the highest M
r
value is observed in the
soils compacted dry of optimum, and the lowest M
r
value
in soils compacted wet of optimum. Although the dry unit
weight of the dry sample is smaller than the OMC sample,
the value of M
r
is higher. This appears to be caused by cap-
illary suction. Capillary suction contributes to increase in
the eective stress by pulling particles towards one another
and thus increasing particle contact force, resulting in
higher M
r
values.
Fig. 5ad are the unconned compressive test results for
OMC, dry and wet samples for I65-146, I65-158, I65-172
and Dsoil, respectively. The unconned compressive
(UC) tests were done to understand why the permanent
strain (which will be discussed in a later section) occurs
excessively for some wet samples, and to understand if
10
12
14
16
18
20
22
24
0.00 5.00 10.00 15.00 20.00 25.00 30.00
Water content, w (%)
D
r
y

u
n
i
t

w
e
i
g
h
t

(
k
N
/
m
3
)
I65-146
I65-158
I65-172
Dsoil
#1soil
#2soil
#3soil
#4soil
SR19
US41
Bloomington
Fig. 3. Compaction curves for soils used.
1474 D. Kim, J.R. Kim / Construction and Building Materials 21 (2007) 14701479
there is any indication of eect of peak strength, stiness of
UC test and permanent strain on resilient behavior. As sta-
ted previously, the repeatability of the UC test was ensured
prior to the main testing. For all of the four sandysilty
clay soils, the highest stiness is observed in the dry sample
and the peak strength is also observed in the dry sample,
except for I65-158 OMC sample. From Fig. 5b and c,
dry samples of I65-158 and Dsoil show slightly larger sti-
ness than OMC samples. The same trend in the repeated
triaxial test is also evidenced in Table 2.
3.4. Siltyclay soils
As shown in Table 2, the seven siltyclay soils have a
slightly dierent resilient behavior compared with the
sandysiltyclay soils. The dierence in M
r
for siltyclay
soils between dry samples and OMC samples are smaller
than that for sandysiltyclay soils. Some OMC samples
show higher M
r
values than dry samples. This indicates
that the eect of dry unit weight on resilient behavior in
the siltyclay soils becomes more pronounced than in the
sandysiltyclay soils and the eect of suction appears to
decrease in the siltyclay soils. Similarly observed in the
sandysiltyclay soils, the wet samples in the siltyclay soils
show considerably lower M
r
values than dry and OMC
samples, which means that the soils are very weak due to
the higher degree of saturation and thus can be used for
the lowest limit (i.e., spring) of M
r
values for subgrades.
3.5. Permanent deformation behavior
Permanent deformation behavior is not considered in
the calculation of M
r
values. This is because the permanent
strain is very small for most of the subgrade soils. For most
of the soils tested, the small permanent deformations
occurred, especially for dry and OMC samples. However,
some samples compacted wet of optimum exhibited an
I65-158 OMC Sample
25000
50000
75000
100000
125000
150000
0.0 30.0 60.0 90.0 120.0
Deviator stress (kPa)
Deviator stress (kPa)
M
r

(
k
P
a
)
M
r

(
k
P
a
)
Conf. = 41.4 kPa
Conf. = 27.6 kPa
Conf. = 13.8 kPa
Conf. = 13.8 kPa
(simplified)
I65-172 OMC Sample
25000
50000
75000
100000
125000
150000
0 30 60 90 120
Conf. = 41.4 kPa
Conf. = 27.6 kPa
Conf. = 13.8 kPa
Conf. = 13.8 kPa
(simplified)
a
b
Fig. 4. Comparison of M
r
(a) between the simplied (500 repetitions for conditioning and 100 repetitions for main testing) and the AASHTO procedures,
(b) between the simplied (250 repetitions for conditioning and 50 repetitions for main testing) and the AASHTO procedures.
Table 2
Measured M
r
values for dry, OMC and wet samples (r
c
= 13.8 kPa,
r
d
= 41.4 kPa)
Soil M
r
values (kPa)
Dry OMC Wet
Sandysiltyclay soils I65-146 94,060 22,940 20,310
I65-158 109,400 76,560 27,370
I65-172 115,210 66,410 17,960
Dsoil 84,660 64,190 13,760
Siltyclay soils #1soil 114,570 86,790 84,560
#2soil 92,690 121,090 16,760
#3soil 99,560 129,720 11,260
#4soil 78,880 73,760 11,840
SR19 172,690 157,870 12,990
US41 166,920 99,900 16,380
Bloomington 94,630 93,000 13,970
D. Kim, J.R. Kim / Construction and Building Materials 21 (2007) 14701479 1475
excessive permanent deformation while performing a
repeated triaxial test. This caused a signicant diculty
to run a repeated triaxial test up to the nal step. Some-
times it was impossible to run a repeated triaxial test to
the end because of the sudden failure of the sample. Most
of the failure was observed to be bulging failure, not shear
failure. As can be seen in Fig. 5ad, the peak strengths of
the wet samples occur at a permanent strain of about seven
percent and the stress ratio of the highest deviator stress
(i.e., 69 kPa) in the repeated triaxial test to the peak
strength are in the range of 5070%. This explains why
the permanent strain occurs excessively in the wet samples.
The AASHTO T307 calls for shear test for samples greater
than 5% permanent strain. However, it is not practical not
to evaluate M
r
values for the soils exhibiting excessive per-
manent deformation. The maximum permanent strain was
set as 20% so that M
r
values can be obtained even for those
soils with excessive permanent strain.
Fig. 6 presents the results of the repeated triaxial test for
I65-146 wet sample in the conditioning stage (step 1) and in
the 5th step, respectively. It was observed in Fig. 6 that
even in the conditioning stage the permanent strain
occurred to about 10% and the permanent strain
approached to about 18% and the testing was terminated
0
50
100
150
200
250
300
A
x
i
a
l

S
t
r
e
s
s

(
k
P
a
)
A
x
i
a
l

S
t
r
e
s
s

(
k
P
a
)
A
x
i
a
l

S
t
r
e
s
s

(
k
P
a
)
A
x
i
a
l

S
t
r
e
s
s

(
k
P
a
)
146omc
146dry
146wet
0
50
100
150
200
250
300
350
0 1 2 3 4 5 6 7 8 9
Axial Strain (%) Axial Strain (%)
Axial Strain (%) Axial Strain (%)
158omc
158dry
158wet
0 1 2 3 4 5 6 7 8
0 1 2 3 4 5 6 7 8 0 1 2 3 4 5 6 7 8
0
50
100
150
200
250
300
172omc
172dry
172wet
0
5
10
15
20
25
30
35
40
dsoil omc
dsoil dry
dsoil wet
a
b
c d
Fig. 5. Unconned compressive test results for dry, OMC, wet samples for (a) I65-146, (b) I65-158, (c) I65-172 and (d) Dsoil.
0
0 1 2 3 4 5 6
3
6
9
12
15
18
21
Step
P
e
r
m
a
n
e
n
t

s
t
r
a
i
n

(
%
)
Fig. 6. Permanent strains for I65-146 wet sample in the conditioning stage
(step 1) in the 5th step.
Permanent strain (9.8%) for wet sample
0
10000
20000
30000
40000
50000
0 20 40 60 80
Deviator stress (kPa)
M
r

(
k
P
a
)
Original length
Deformed length
Fig. 7. M
r
values for original length and deformed length.
1476 D. Kim, J.R. Kim / Construction and Building Materials 21 (2007) 14701479
in the 5th step. A comparison was made of the resilient
modulus between using the original length and using the
deformed length for I65-158 soil, as shown in Fig. 7. The
permanent strain of about 10% occurred in the repeated tri-
axial test and the dierence in M
r
values using the original
and deformed lengths are approximately eight percent.
This implies that it would be more accurate to calculate
the M
r
values using the deformed length.
3.6. A predictive model for M
r
values
Generally, resilient behavior of subgrade soils can be
described by the Uzan model [17] known as the universal
model, taking into account conning and deviator stresses.
The predicted M
r
values can be obtained from the follow-
ing equation:
M
r
k
1
p
a
h
p
a

k
2
r
d
p
a

k
3
2
where k
1,
k
2,
k
3,
is the regression coecients; h the sum of
principal stresses; p
a
the reference pressure 100 kPa
1 kgf/cm
2
2000 psf 14.5 psi; and r
d
is the deviator
stress in the same unit as p
a
.
As noticed in Eq. (2), the major limitation of Uzans
model is that it cannot take into account the dierent types
of subgrade soils [18]. In order to overcome such limita-
tion, a predictive model based on the all the repeated triax-
ial test data for 11 soils, as shown in Table 3, the material
coecients k
1
, k
2
, and k
3
were developed through multiple
regression analyses in terms of 11 soil variables, which can
be easily obtained in the sieve analysis, Atterbergs Limit
test, and standard Proctor compaction test. As shown in
Table 3
Predictive model for estimation of M
r
values considering 11 soil variables
log k
1
= 6.660876 0.22136 OMC 0.04437 MC 0.92743 MCR 0.06133 DD + 10.64862 %comp + 0.328465 SATU 0.04434
%sand 0.04349 %SILT 0.01832 %Clay + 0.027832 LL 0.01665 PI (R
2
= 0.85)
k
2
= 3.952635 0.33897 OMC + 0.076116 MC 2.45921 MCR 0.06462 DD + 6.012966 %comp + 1.559769 SATU + 0.020286 %
sand + 0.002321 %SILT + 0.011056 % CLAY + 0.077436 LL 0.05367 PI (R
2
= 0.81)
k
3
= 2.634084 + 0.124471 OMC 0.09277 MC + 0.366778 MCR 0.01168 DD 1.32637 %comp + 1.297904 SATU 0.01226 %
sand 0.00512 % SILT 0.00492 %clay 0.05083 LL + 0.018864 PI (R
2
= 0.76)
y = 0.7908x + 118.04
R
2
= 0.8338
0
400
800
1200
1600
2000
0 400 800 1200 1600 2000
Measured k1
P
r
e
d
i
c
t
e
d

k
1
y = 0.7458x + 0.0699
R
2
= 0.7466
-1
-0.5
0
0.5
1
1.5
-1 -0.5 0 0.5 1 1.5
Measured k2
P
r
e
d
i
c
t
e
d

k
2
y = 1.1365x - 0.4071
R
2
= 0.5136
-1.4
-0.9
-0.4
0.1
0.6
-1.4 -0.9 -0.4 0.1 0.6
Measured k3
P
r
e
d
i
c
t
e
d

k
3y = 1.1365x
R
2
= 0.8472
0
50000
100000
150000
200000
0 50000 100000 150000 200000
Measured Mr (kPa)
P
r
e
d
i
c
t
e
d

M
r

(
k
P
a
)
a
b
c d
Fig. 8. Comparison between (a) predicted k
1
and measured k
1
, (b) predicted k
2
and measured k
2
, (c) predicted k
3
and measured k
3
, (d) predicted M
r
and
measured M
r
values.
D. Kim, J.R. Kim / Construction and Building Materials 21 (2007) 14701479 1477
Table 3, the coecient of determination, R
2
, for k
1
, k
2
, and
k
3
are reasonable to estimate the resilient modulus once the
soil variables are known. The variables are the following:
OMC, MC (moisture content), MCR (moisture content
ratio = moisture content/optimum moisture content), DD
(dry density), SATU (degree of saturation), % sand (per-
cent sand in particle size distribution curve), % CLAY (per-
cent sand in particle size distribution curve), LL (liquid
limit) and PI (plasticity index).
Fig. 8ac are the comparisons between measured and
predicted regression coecients and Fig. 8d is the compar-
ison between measured and predicted resilient values for
675 data points for 45 samples tested. Predicted regression
coecients and resilient modulus values are reasonably
comparable to measured ones considering the wide range
of water contents adopted in this study (Se/Sy = 0.45,
R
2
= 0.85). This predictive model can be readily used in
the estimation of M
r
values in design phase of road
construction.
4. Conclusions
In this study, four sandysiltyclay and seven siltyclay
Indiana subgrade soils were used to evaluate resilient
behavior of these soils. The following can be drawn from
this study.
(1) The simplied procedure suggested compares well
with the existing repeated triaxial test procedure. This
can signicantly reduce the time required for the
repeated triaxial test with reasonable accuracy.
(2) For some soils, excessive permanent strains occurred
while testing. This is because the stress ratio of the
deviator stress to the peak strength of those soils were
quite large and permanent strain to reach the peak
strength is quite large, approximately 7%.
(3) The largest M
r
values are observed in the dry samples
for sandysiltyclay soils due to the capillary suction
while the largest M
r
values are observed either in the
dry or OMC sample for siltyclay soils. The smallest
M
r
values obtained from wet samples can be used as
the limit of M
r
in spring.
(4) The current repeated triaxial test uses the original
length of the specimen, but the deformed length dur-
ing testing should be used for more accurate calcula-
tion of M
r
.
(5) A predictive model was developed using the 11 vari-
ables easily obtained from physical property tests
and the standard Proctor test. The predicted regres-
sion coecients and resilient moduli compare well
with the measured ones.
5. Recommendations
The current repeated triaxial test cannot take into
account the long term M
r
values due to a limited number
of repeated loadings applied to the specimen. The long
term M
r
values are especially needed for rehabilitation. A
study on the long term M
r
values is recommended.
Acknowledgements
The authors are grateful to the Federal Highway
Administration/Indiana Department of Transportation/
Joint Transportation Research Project (Project: SPR
2633) for supporting this research. They are obliged to
the Study Advisory Committee members: Vincent
Drnevich (Purdue); Kumar Dave, Samy Noureldin
and Nayyar Zia (INDOT); Val Straumins (FHWA).
Also, appreciation is extended to Tommy Nantung of
INDOT Research Division for his support and encour-
agement.
References
[1] Mohammad LN, Titi HH, Herath A. Intrusion technology: An
innovative approach to evaluate resilient modulus of subgrade
soils, Geotechnical Special Publication No. 85, ASCE; 1998. p. 39
58.
[2] Elliot RP, Thornton SI. Simplication of subgrade resilient modulus
testing. Transportation Research Record, 1192, TRB. Washington
DC: National Research Council, 1988. p. 17.
[3] Elliot RP, Thornton SI. Resilient modulus and AASHTO pavement
design. Transportation research record, 1196, TRB. Washington DC:
National Research Council, 1988. p. 17.
[4] Ping WV, Yang Z, Liu C, Dietrich B. Measuring resilient modulus of
granular materials in exible pavements, transportation research
record, 1778, TRB. National Research Council, Washington DC,
2001. p. 8190.
[5] Drumm EC, Reeves JS, Madgett MR, Trolinger WD. Estimation of
subgrade resilient modulus from standard tests. J Geotech Geoenvi-
ron Eng ASCE 1997;123(7):66370.
[6] Drumm EC, Boateng-Poku Y, Pierce TJ. Estimation of subgrade
resilient modulus from standard tests. J Geotech Geoenviron Eng
ASCE 1990;116(5):77589.
[7] Lee W, Bohra NC, Altschae AG. Subgrade resilient
modulus for pavement design and evaluation, Final report,
JHRP-92-23, Purdue University, Lafayette W. (Ed.), IN;
1993.
[8] Kim D, Kweon G, Lee K. Alternative method of determining resilient
modulus of subgrade soils using a static triaxial test. Cand Geotechn J
2001;37:10716.
[9] Houston WN, Houston SL, Anderson TW. Stress state consideration
for resilient modulus testing of pavement subgrade, Transportation
Research Record, 1406, TRB. National Research Council, Washing-
ton DC: 1993. p. 12432.
[10] Altschae AG, Duckworth RA, Clough MK. Implementation of
subgrade resilient modulus for pavement design and evaluation. Final
report, FHWA/IN/JTRP-98/2, Purdue University, Lafayette W.
(Ed.), IN: 1998.
[11] Santa BL. Resilient modulus of subgrade soils: comparison of two
constitutive equations, Transportation Research Record, 1462,
TRB. Washington DC: National Research Council, 1994. p. 79
90.
[12] Thompson MR, Robnett QL. Resilient properties of subgrade soils.
Proc ASCE 1979;105(TE1):7189.
[13] Wilson BE, Sargand SM, Hazen GA, Green R. Multiaxial
testing of subgrade, Transportation Research Record, 1278,
TRB. Washington DC: National Research Council, 1990. p.
915.
1478 D. Kim, J.R. Kim / Construction and Building Materials 21 (2007) 14701479
[14] Lee W, Altschae AG, White TD. Resilient modulus of cohesive
soils. J Geotech Geoenviron Eng ASCE 1997;123(2):1316.
[15] Lee W, Bohra NC, Altschae AG. Resilient characteristics of dune
sand. J Trans Eng ASCE 1995;12(6):5026.
[16] Kim D, Salgado R, Altschaee AG. Eects of super-single tire
loadings on subgrades. J Trans Eng ASCE 2005(October):73243.
[17] Uzan J. Characterization of granular material. Transportation
Research Record, 1022, TRB. Washington DC: National Research
Council, 1985. p. 529.
[18] Yau A, Quintus HL, Study of LTPP laboratory resilient modulus test
data and response characteristics, Final report, FHWA-RD-02-051,
2002.
D. Kim, J.R. Kim / Construction and Building Materials 21 (2007) 14701479 1479