See discussions, stats, and author profiles for this publication at: https://www.researchgate.

net/publication/239439061

Evaluation of Subgrade Resilient Modulus Predictive Model for Use in

Mechanistic-Empirical Pavement Design Guide

Article  in  Transportation Research Record Journal of the Transportation Research Board · January 2006
DOI: 10.3141/1947-15

CITATIONS READS

26 434

5 authors, including:

Lev Khazanovich Clara Celauro

University of Pittsburgh Università degli Studi di Palermo
136 PUBLICATIONS   1,232 CITATIONS    47 PUBLICATIONS   547 CITATIONS   

SEE PROFILE SEE PROFILE

Some of the authors of this publication are also working on these related projects:

Special Issue "Sustainability of Soil Reuse in Civil Construction" View project

Research Project - PRIN: PROGETTI DI RICERCA DI RILEVANTE INTERESSE NAZIONALE – Bando 2017 - Prot. 2017XYM8KC “Urban safety, sustainability, and resilience: 3 paving
solutions, 4 sets of modules, 2 platforms.” Acronym: USR342. View project

All content following this page was uploaded by Lev Khazanovich on 06 February 2016.

The user has requested enhancement of the downloaded file.

Evaluation of Subgrade Resilient
Modulus Predictive Model for Use
in Mechanistic–Empirical Pavement
Design Guide
Lev Khazanovich, Clara Celauro, Bruce Chadbourn, John Zollars,
and Shongtao Dai
The characterization of unbound materials in the mechanistic–empirical
pavement design guide (MEPDG), also known as the 2002 design guide,
is reviewed, and this characterization is applied to Minnesota subgrades.
The main emphasis is on the collection of k1-, k2-, and k3-parameters for
Minnesota fine-grained soils and the procedure for the interpretation of
the resilient modulus test to provide an input to the multilayer elastic
theory (MLET) analysis (Level 2 input). This is an important aspect of
adaptation of the MEPDG, because the guide recommends measurement
of resilient moduli from laboratory testing, but the procedure does not
specify how to interpret the test data to obtain an input for an MLET
analysis. The resilient modulus test results from 23 samples collected from
several Minnesota locations were used to provide information for the
nonlinear finite element program and multilayer elastic theory program.
The obtained elastic moduli were compared with the MEPDG recom-
mended ranges for subgrade modulus of elasticity based on the soil clas-
sification (Level 3 inputs). The MEPDG Level 3 ranges were found to be
reasonable.
A new mechanistic-based design procedure—the mechanistic–
empirical pavement design guide (MEPDG), also known as the
2002 design guide—has been recently developed by NCHRP under
AASHTO sponsorship (1). Many state transportation agencies, includ-
ing the Minnesota Department of Transportation (Mn/DOT), began
the evaluation of this procedure with the ultimate goal of its adapta-
tion and calibration for local conditions. An important aspect of this
process is evaluation of the required inputs.
Design inputs include traffic (full-load spectra for single, tandem,
tridem, and quad axles), material and subgrade characterization,
climatic factors, performance criteria, shoulders, subdrainage, joint
details, and many others. One of the innovative aspects of the design
procedure is the consideration of different levels of inputs. Level 1
requires the engineer to obtain the most accurate design inputs
[e.g., direct testing of the materials, on-site weigh-in-motion (WIM)
L. Khazanovich and C. Celauro, Department of Civil Engineering, University of Min-
nesota, Minneapolis, MN 55455. Current affiliation for C. Celauro: Dipartimento
di Ingegneria delle Infrastrutture Viarie, Università degli Studi di Palermo, Italy.
B. Chadbourn, J. Zollars, and S. Dai, Office of Materials, Minnesota Department
of Transportation, 1400 Gervais Avenue, Maplewood, MN 55109.
Transportation Research Record: Journal of the Transportation Research Board,
No. 1947, Transportation Research Board of the National Academies, Washington,
D.C., 2006, pp. 155–166.
155
data]. Level 2 requires testing, but the use of correlations is allowed
(for example, subgrade modulus estimated through correlation with
another test), and Level 3 generally uses estimated values. Thus,
Level 1 has the least possible error associated with inputs, Level 2
has more error, and Level 3 has the most.
The hierarchical approach to design inputs provides the designer
with flexibility in obtaining the design inputs for a design project
based on the importance of the project and the available resources
(1). For a given design project, inputs may be obtained by using a
mix of levels, such as traffic load spectra from Level 1, asphalt layer
characterization from Level 2, and subgrade resilient modulus
from Level 3.
For the majority of the input parameters, regardless of input
design levels, the computational algorithm for damage is exactly
the same. The same models and procedures are used to predict
distress and smoothness no matter what levels are used to obtain
the design inputs. This is not the case for characterization of unbound
materials, for which the input level affects the procedure for cal-
culation of structural responses of the pavement system (strains,
stresses, and deflections). The experience gained in an evaluation
of resilient modulus input for Minnesota subgrades for use in the
MEPDG is shared in this paper.
CHARACTERIZATION OF UNBOUND
MATERIALS IN MEPDG
The MEPDG requires the provision of resilient moduli for all
unbound layers and subgrade for each design period (month) (1, 2).
The guide recommends the measurement of resilient moduli from
the laboratory, obtaining them through the use of correlations with
other material properties or estimating them on the basis of soil
classification. Three levels of inputs are provided for characteri-
zation of resilient properties of unbound materials at the optimum
moisture content.
Level 1. Laboratory Testing
Laboratory testing of resilient moduli is required for Level 1 input.
The resilient modulus laboratory testing involves cyclic triaxial
testing of prepared representative samples of unbound materials or
subgrade. The design guide recommends the following standard test
156 Transportation Research Record 1947

methods for the laboratory preparation, testing, and computation of • Provide the resilient modulus for each design period and
the test results: • Provide the resilient modulus for the optimum moisture content.

• NCHRP Project 1-28: Harmonized Test Methods for Laboratory
Determination of Resilient Modulus for Flexible Pavement Design
(3) and
• AASHTO T 307: Determining the Resilient Modulus of Soil
and Aggregate Materials.
If the second option is selected, the enhanced Integrated Climatic
Model incorporated into the MEPDG software predicts seasonal
variation in the moisture content of the unbound layers (6 ). Then
the MEPDG software adjusts the moduli for the other moisture
conditions by using the following model (7, 8):

For mechanistic–empirical (M-E) design, the resilient modulus is a+

b−a
1+ exp[β + kS i ( S − S0 )]
estimated by using the following generalized k1-k2-k3 constitutive M R = 10 M Ropt (2)
model:
where
k2 k3
⎛ θ⎞ ⎛ τ oct ⎞
M Ropt = k1 i i
pa ⎜ ⎟ i
⎜⎝ p + 1⎟⎠ (1) MR = resilient modulus at any degree of saturation;
⎝ pa ⎠ a S = degree of saturation while material is being tested;
MRopt = resilient modulus at optimum water content and maximum
where dry density;
MRopt = resilient modulus at optimum moisture content, S0 = degree of saturation at optimum water content;
k1, k2, k3 = regression parameters, a = minimum of log (MR/MRopt);
pa = atmospheric (normalizing) pressure, b = maximum of log (MR/MRopt);
θ = J1 = σ1 + σ2 + σ3 = bulk stress (first stress invariant), and β = location parameter, obtained as function of a and b; and
1 kS = regression parameter.
τoct = ( σ 1 − σ 2 )2 + ( σ 1 − σ 3 )2 + ( σ 2 − σ 3 )2
3 For fine-grained materials, the guide recommends the following
= octahedral shear stress. model parameters: a = −0.5934, b = 0.4, β = −0.3944, kS = 6.1324.
The k1-k2-k3 model is directly applicable to design only if Level 1 for Figure 1 presents the correction factor for the moisture condition for
subgrade is selected and the nonlinear axisymmetric finite element the various degrees of saturation. One can observe that an increase
program DSC-2D is used to compute the structural responses of in moisture content decreases the resilient modulus.
flexible pavements (4 ). Substitution of Equation 1 into Equation 2 leads to the following
predictive model for the resilient model under an arbitrary moisture
condition:
Level 2. Correlations with Other
Material Properties b−a k2 k3
a+
1+ exp[β + kS i ( S − S0 )] ⎛ θ⎞ ⎛ τ oct ⎞
M R = 10 k1 i pa i ⎜ ⎟ i
⎜⎝ p + 1⎟⎠ (3)
Level 2 inputs for unbound layers and subgrade should be used if ⎝ pa ⎠ a
the multilayer elastic theory (MLET) program JULEA is used for
calculation of stresses, strains, and deflections in flexible pavements Equation 3 can be rewritten in the following form:
(5). Laboratory testing is still the preferable source of information
k2 k3
for this level of testing. However, if no resilient modulus laboratory ⎛ θ⎞ ⎛ τ oct ⎞
M R = k *1 i pa i ⎜ ⎟ i
⎜⎝ p + 1⎟⎠ (4)
test data are available, the MEPDG software calculates the resilient ⎝ pa ⎠ a
modulus by using the relationships with other properties, such as
California bearing ratio or R-value. If the laboratory data or param- where
eters of the k1-k2-k3 model are available, the design guide recommends
an iterative procedure, which would result in a single representative b−a
a+
value for the resilient modulus. That procedure has been refined in this 1+ exp[β + kS i ( S − S0 )]
k *1 = 10 k1 (5)
study. The discussion of conversion of the parameters of the k1-k2-k3
model is presented in the next section of this paper. One can see that the MEPDG moisture correction model postulates
that change in moisture conditions affects only the k1-parameters but
Level 3. Typical Values does not affect k2 and k3.

For input Level 3, the resilient modulus for the optimum moisture
content is selected on the basis of the material classification. The
MLET analysis should be used if the Level 3 input option is selected.
In this study, the national defaults for the fine-grained soils were
evaluated. The results of the evaluation are presented below.
SEASONAL VARIATION IN RESILIENT MODULI
The MEPDG permits accounting for seasonal variation in properties
of unbound materials by adjustment of the resilient modulus for each
design period (month). The user has two options:
DATABASE OF k1-k2-k3 MODEL PARAMETERS
Evaluation of typical inputs required by the MEPDG is an important
step toward its implementation. In this study, the results of a variety of
resilient modulus tests conducted by the MnDOT Office of Materials
between October 2000 and February 2002 on Minnesota fine sub-
grades were evaluated, and a database of parameters k1, k2, and k3
was created. The majority of the subgrades are A-6 according to the
AASHTO classification, although several specimens represent A-7-6
and A-7-5 subgrades. The tests were conducted according to the
Long-Term Pavement Performance (LTPP) Protocol 46 for mea-
Khazanovich, Celauro, Chadbourn, Zollars, and Dai 157

3.00

2.50

2.00
Correction Factor

1.50

1.00

0.50

0.00
-2 -1.5 -1 -0.5 0 0.5 1 1.5 2 2.5 3
S - Sopt

FIGURE 1 Correction factor as function of degree of saturation.

suring the resilient modulus of unbound granular base and subbase
materials and subgrade soils. Since not all the tests were carried out
at the optimum moisture content, the coefficient k1 was obtained
with Equation 4. Table 1 presents the resulting parameters k1, k*1,
k2, and k3 obtained from the regression analysis along with the
moisture content of the subgrade during the test.
Analysis of parameters k1, k2, and k3 from Table 1 shows that even
for subgrades located in one state and having the same class according
to the AASHTO classification, these parameters may vary signifi-
cantly. It was not found possible to come up with typical values for
these parameters. Instead, the following ranges were proposed for
Minnesota fine-grained subgrade:
with the Level 1 input. If the MLET analysis (Level 2 input) is used,
only a single value of the modulus of elasticity should be provided.
The resilient modulus selection should take into account the stress
state induced in the unbound layer by a typical wheel load and at-rest
earth pressure. In contrast, the stress distribution in the multilayer
system is affected by the modulus of elasticity of the individual
layers. Therefore, an iterative procedure should be used for the
modulus calculation.
The following approach was proposed several decades ago by
Witczak and Smith and, with minor modifications, adopted by the
MEPDG (9). The recommended procedure involves the following
steps:

• k1: from 1,000 to 5,000; Step 1. Assume initial moduli for each unbound layer in the pave-
• k2: from 0.01 to 0.35; and ment system.
• k3: from −6 to −1.5. Step 2. Compute a stress state for critical points within unbound
layers from the wheel load.
These ranges may be used for an initial evaluation of quality of Step 3. Use the stresses obtained in Step 2 to compute the total
the resilient modulus test data. If the resulting coefficients k1, k2, and stress state from the wheel load and the overburden pressure. The
k3 fall outside these ranges, the test should be repeated to confirm vertical component of the stress state, σZ, should be computed as
the measurements. follows:

σ Z = σ V + p0 (6)
DETERMINATION OF MEPDG LEVEL 2
SUBGRADE INPUT FROM K1-K2-K3 MODEL where σV is the vertical stress from the wheel load as computed with
elastic layer theory and p0 is the at-rest vertical pressure from the
General Procedure
overburden of other layers:
The MEPDG strongly encourages laboratory testing of unbound n −1
materials to determine parameters k1, k2, and k3. These parameters, p0 = hn i γ n + ∑ hi i γ i (7)
however, can be used as a direct input into the MEPDG software only i =1
158 Transportation Research Record 1947

TABLE 1 Parameters k 1 -k 2 -k 3 for Minnesota Subgrades

Dry Optimum Test

AASHTO Density, Moisture Moisture
Location Class lb/ft3 Content, % Content, % k*1 k1 k2 k3

City of Rochester A-7-6 106.27 18 16.2 2554.302 1815.049 0.180793 −4.23504

City of Rochester A-6 104.22 19.2 17.71 1547.234 1181.946 0.160933 −2.45363
Isanti County A-6 111.46 12.9 11.4 4625.851 3168.056 0.165225 −4.23921
Isanti County A-6 109.81 12.9 13.75 2314.683 2883.263 0.265045 −4.15
Isanti County A-6 110.57 12.9 11.95 2050.835 1596.169 0.18395 −2.35394
Itasca County A-6 111.77 15.1 14.79 2310.489 2149.032 0.113406 −3.29396
Kandiyohi County A-6 114.57 13.1 11.87 2899.871 2072.444 0.249715 −3.09559
Kandiyohi County A-6 115.25 13.1 12.18 2438.324 1906.016 0.204286 −2.95031
Lyon County A-6 112.65 14.5 14.11 1821.373 1663.879 0.233177 −4.49975
Lyon County A-6 116.34 12.5 11.74 3636.023 2955.099 0.162358 −3.56106
Medford Roundabout A-6 116.55 13.2 12.37 2087.053 1670.209 0.250131 −2.82977
MnRoad Cell 33 A-6 115.44 15 15.42 2989.392 3277.297 0.156599 −5.13509
MnRoad Cell 34 A-6 119.4 15 12.91 2317.111 1532.798 0.156738 −2.64929
MnRoad Cell 35 A-6 118.79 15 13.68 2618.802 2001.015 0.232257 −4.3486
St. Louis County A-7-6 90.22 25.6 27.63 1266.482 1661.842 0.02788 −3.04364
St. Louis County A-7-5 89.41 23.7 23.07 1596.688 1437.535 0.012749 −1.91293
Stearns County A-6 104.66 17.8 15.69 1190.627 798.2841 0.323623 −3.42635
Stearns County A-6 110.99 14.3 13.32 2963.512 2296.077 0.321999 −5.8838
Wright County A-7-6 97.9 21.3 21.44 1256.837 1285.827 0.207286 −5.05057
Wright County A-6 106.75 21 19.05 1109.774 829.0605 0.290016 −5.33975
Wright County A-6 103.61 21 19.9 674.3569 563.9285 0.377793 −4.96021
Wright County A-7-6 106.86 16.5 17.97 1369.223 1861.696 0.21318 −4.04342
Wright County A-7-6 108.82 16.5 15.6 2496.026 2064.54 0.054135 −1.69998

where
i = layer above the soil layer, n, for which the resilient modulus is
being estimated—it is assumed here that Layer 1 is the surface
and that i increases with depth,
hi = thickness of the ith layer or depth from the surface of the nth
layer to the stress calculation point of interest, and
γi = density of the ith layer.
For computing the contribution of the overburden stresses to
the total lateral stress, it is necessary to estimate the at-rest pres-
sure coefficient, K0. Such a coefficient for cohesive soils (such as
clays) is normally calculated as a function of the Poisson’s ratio, µ,
according to
Step 4. Using the stress obtained in Step 3 and the constitutive
Equation 1, compute the predicted resilient moduli.
Step 5. For each unbound layer, compare the resilient moduli
computed in Step 4 with the assumed moduli. If these moduli sig-
nificantly differ from each other, Steps 2 through 4 should be repeated
with the set of the moduli obtained in the last iteration.
The foregoing procedure closely follows the guidelines provided
in the MEPDG. However, its implementation is not straightforward
since many important details are not specified in the guide. After con-
sultations with the NCHRP Project 1-40B team (H. L. Von Quintus,
personal communication), the following refinements of the MEPDG
procedure were initially adopted in this study:

• The iteration process was assumed to be converged if the relative

μ
K0 = (8) errors between assumed and calculated moduli were less than 1%
1− μ for each unbound layer.
• The wheel load used for computation of the traffic stresses was
For noncohesive soils (such as gravel and sand), the at-rest pressure assumed to be a falling weight deflectometer (FWD)–type single-
coefficient is a function of the angle of shearing resistance, ϕ: wheel 9,000-lb load.
• The following locations were specified for computing stresses
K 0 = 1 − sin ϕ (9) in the unbound layers:
– Each unbound layer except the last one: one-fourth of the
with ϕ measured in radians. depth,
The total lateral stress, σ3, for determining the resilient modulus – The last layer: 18 in. from the top surface.
of the soil layer then can be determined from

σ 3 = σ H + K 0 i p0 (10) Procedure Implementation

where σH is the horizontal stress from the wheel load applied at the Although the procedure just described is relatively simple, its imple-
pavement surface and computed with elastic layer theory. mentation is time-consuming because all iterations require perform-
Khazanovich, Celauro, Chadbourn, Zollars, and Dai
ing an MLET analysis run. To speed up the process, several rapid
solutions were developed for predicting critical MLET responses for
typical pavement systems for the locations specified in the procedure
described earlier.
A factorial of the MLET program was performed for a four-layer
system loaded by an FWD-type load as shown in Figure 2. The top
layer simulated an asphalt layer; the second and the third layers from
the top simulated a granular base and subbase, respectively; and the
semi-infinite layer modeled a subgrade. The stresses were calculated
in the second and third layers at one-fourth of the depth (Positions
A and B) and at 18 in. below the top of the subgrade (Position C).
Almost 600,000 MLET program runs were performed to create
a training database. The Poisson’s ratios of the layers were kept
constant and assumed to be equal to 0.3, 0.35, 0.35, and 0.45 for the
asphalt, base, subbase, and subgrade layers, respectively. The layer
thicknesses and moduli of elasticity varied in wide ranges to
cover the typical values of these parameters. The following values
for the thickness and the modulus of elasticity were considered:
• Asphalt concrete (AC) thickness: 2, 4, 6, 8, 10, and 12 in.;
• Base thickness: 6, 9, 12, 15, 18, and 21 in.;
• Subbase thickness: 6, 9, 12, 15, 18, and 21 in.;
• AC modulus of elasticity: 200, 400, 600, 800, 1,000, 1,500, and
2,000 ksi;
• Base modulus of elasticity: 10, 15, 20, 25, 30, and 40 ksi;
• Subbase modulus of elasticity: 10, 15, 20, 25, 30, and 40 ksi; and
• Subgrade modulus of elasticity: 3, 5, 7, 9, 11, 13, 15, 18, 22,
25, and 30 ksi.
On the basis of the results of these runs, several rapid solutions
were developed. They permitted computations of the stresses at the
critical locations much faster than would be done with the MLET
program. More important, these solutions could be called directly
HMA Layer
Pos. A Granular Base
Pos. B
Granular Subbase
Subgrade
18 in.
Pos. C
FIGURE 2 Pavement system and evaluation points
used for stress state analysis.
159
from the Excel spreadsheet used to implement the procedure for
resilient modulus iterations. This method permitted an efficient
analysis of the effect of the unbound layer parameters and asphalt
layer thickness on predicted resilient moduli.
SENSITIVITY ANALYSIS
With the rapid solutions described, several sensitivity studies
were conducted to investigate the effect of parameters k1, k2, and k3
on the predicted resilient moduli. The results of one such study are
presented next.
A series of runs was performed for the following baseline
parameters:
• AC thickness and modulus of elasticity were assumed to be
equal to 4 in. and 600,000 psi, respectively;
• Base and subbase thicknesses were assumed to be equal to 6 in.;
• Base k1-, k2-, and k3-parameters were assumed to be equal to
14,323, 0.758, and −0.288, respectively; these parameters were
reported by Witczak et al. for one of the base materials in the Arizona
Department of Transportation database;
• Subbase k1-, k2-, and k3-parameters were assumed to be equal
to 1,888.9, 0.5306, and −1.653, respectively; and
• Subgrade k1-, k2-, and k3-parameters were assumed to be equal
to 1,468, 0.838, and −0.888, respectively.
For the selected pavement structure, one parameter was changed
and the iterative procedure for predicting the resilient modulus was
performed. Several examples of this analysis are presented next.
Figure 3 presents the effect of the AC, base, and subbase layer
thicknesses on the predicted resilient moduli. Figure 3a shows that
an increase in asphalt thickness significantly reduces the predicted
base modulus but has a less pronounced effect on the subbase and
subgrade moduli. Nevertheless, an increase in AC thickness from 2
to 8 in. leads to an increase of the subgrade resilient modulus from
11.9 to 13.7 ksi. A similar effect was observed for the base thick-
ness (see Figure 3b). Although the change in the base resilient
modulus was not as significant as was observed when the AC thick-
ness was changed, the subgrade resilient modulus changed from
12.5 to 14.1 ksi when the base thickness increased from 6 to 21 in.
As can be observed from Figure 3c, an increase in subbase thickness
does not substantially change the base and subbase moduli but
increases the subgrade modulus.
Naturally, the material model parameters (k1, k2, and k3) affect the
predicted moduli. Figure 4 presents the effect of parameters k1, k2,
and k3, respectively, of the base layer on the predicted resilient
moduli in the unbound layers. One can observe that an increase in
these parameters increases the predicted resilient modulus of the
base layer but does not significantly change the resilient moduli of
the subbase and subgrade.
Figure 5 shows that a change in parameters k1, k2, and k3 of the
subbase layer may affect not only the resilient modulus of the sub-
base but also the resilient modulus of the base layer. At the same
time, no appreciable effect of change in subbase k1, k2, and k3 on the
subgrade resilient modulus was observed.
The effects of the k1, k2, and k3 of the subgrade on the pavement
system resilient moduli are shown in Figure 6. It can be observed
from Figure 6a that parameter k1 significantly affects not only the
resilient modulus of the subgrade but also the resilient moduli of the
base and subgrade. The effects of parameters k2 and k3 are much less
pronounced.
160 Transportation Research Record 1947

40,000
E Base
35,000 E Subbase
E Subgrade

Resilient Modulus Mr (psi)

30,000

25,000

20,000

15,000

10,000

5,000

0
0 2 4 6 8 10
Asphalt Thickness (in.)
(a)

40,000

35,000 E Base
E Subbase
Resilient Modulus Mr (psi)

30,000 E Subgrade

25,000

20,000

15,000

10,000

5,000

0
0 5 10 15 20 25
Base Thickness (in.)
(b)

40,000
E Base
35,000
E Subbase
Resilient Modulus Mr (psi)

30,000 E Subgrade

25,000

20,000

15,000

10,000

5,000

0
0 5 10 15 20 25
Subbase Thickness (in.)
(c)

FIGURE 3 Effect of AC, base, and subbase layer thicknesses on predicted resilient
moduli.
Khazanovich, Celauro, Chadbourn, Zollars, and Dai 161

40,000

E Base
35,000
E Subbase

Resilient Modulus Mr (psi)

30,000 E Subgrade

25,000

20,000

15,000

10,000

5,000

0
600 900 1,200 1,500
k1 Base
(a)

40,000

35,000
E Base
Resilient Modulus Mr (psi)

30,000 E Subbase
E Subgrade
25,000

20,000

15,000

10,000

5,000

0
0 0.2 0.4 0.6 0.8 1 1.2
k2 Base
(b)

40,000
E Base
35,000
E Subbase
Resilient Modulus Mr (psi)

30,000
E Subgrade
25,000

20,000

15,000

10,000

5,000

0
–1.00 –0.80 –0.60 –0.40 –0.20 0.00
k3 Base
(c)

FIGURE 4 Effect of base parameters k 1 , k 2 , and k 3 on predicted resilient moduli.

162 Transportation Research Record 1947

40,000

35,000 E Base
E Subbase

Resilient Modulus Mr (psi)

30,000
E Subgrade
25,000

20,000

15,000

10,000

5,000

0
600 1,100 1,600 2,100 2,600
k1 Subbase
(a)

40,000
E Base
35,000
E Subbase
Resilient Modulus Mr (psi)

30,000 E Subgrade

25,000

20,000

15,000

10,000

5,000

0
0 0.2 0.4 0.6 0.8 1 1.2
k2 Subbase
(b)

40,000
E Base
35,000
E Subbase
Resilient Modulus Mr (psi)

30,000 E Subgrade

25,000

20,000

15,000

10,000

5,000

0
–2.00 –1.80 –1.60 –1.40 –1.20 –1.00
k3 Subbase
(c)

FIGURE 5 Effect of subbase parameters k 1 , k 2 , and k 3 on predicted resilient moduli.

Khazanovich, Celauro, Chadbourn, Zollars, and Dai 163

40,000
E Base
35,000 E Subbase
E Subgrade

Resilient Modulus Mr (psi)

30,000

25,000

20,000

15,000

10,000

5,000

0
1,000 1,500 2,000 2,500 3,000 3,500
k1 Subgrade
(a)

40,000

35,000 E Base
E Subbase
Resilient Modulus Mr (psi)

30,000 E Subgrade

25,000

20,000

15,000

10,000

5,000

0
0.00E+00 1.00E-02 2.00E-02 3.00E-02 4.00E-02 5.00E-02 6.00E-02
k2 Subgrade
(b)

40,000
E Base
35,000
E Subbase
Resilient Modulus Mr (psi)

30,000 E Subgrade

25,000

20,000

15,000

10,000

5,000

0
–4.00 –3.50 –3.00 –2.50 –2.00 –1.50
k3 Subgrade
(c)

FIGURE 6 Effect of subgrade parameters k 1 , k 2 , and k 3 on predicted resilient moduli.

164 Transportation Research Record 1947

EVALUATION OF LEVEL 2 RESILIENT MODULI
FOR MINNESOTA SUBGRADES
To evaluate the robustness of the conversion of the k1-k2-k3 model
into a single resilient modulus, this procedure was applied to the
Minnesota subgrades (see Table 1) for a variety of hypothetical
pavement structures. The following combination of AC and base
thicknesses was considered:
• AC thickness: 2, 4, 6, 8, 10, and 12 in.;
• AC modulus of elasticity: 200,000 and 2,000,000 psi;
• Base thickness: 6, 9, and 12 in.;
• Base parameters k1, k2, and k3: 14,323, 0.758, and −0.288,
respectively; and
• Subbase thickness: 8 in.
Two levels of AC modulus of elasticity represent very stiff and very
soft AC layers. The base layer was assumed to have properties of the
Grey Mountain granular base material from the Arizona database
(7 ). The subbase layer was assumed to have the same properties as
the subgrade, but the minimum resilient modulus was set to be equal
to 10,000 psi. The maximum modulus of elasticity was assumed
to be equal to 40,000 psi for the base layer and 30,000 psi for the
subbase and subgrade.
For each subgrade type and pavement structure (total of 36 cases
for each subgrade), a resilient modulus was calculated with the
modified iterative procedure. Table 2 presents the summaries of
the subgrade and subbase resilient moduli in terms of ranges of the
observed values and the mean value for each subgrade type. Also
shown are the recommended ranges based on the AASHTO material
classification.
Analysis of Table 2 shows that, with a few exceptions, the MEPDG
Level 3 recommendations for the subgrade resilient modulus repre-
sent a reasonable estimate of the resilient modulus. At the same time,
even for the subgrade of the same AASHTO soil classification class
located in the same state, the resilient modulus may vary significantly.
Moreover, even for the same material, the ranges of predicted resilient
moduli can be quite wide. The subgrade resilient modulus would be
much higher for a pavement system with a thick and stiff asphalt
layer and much lower for a pavement system with a thin and soft
asphalt layer. Therefore, assignment of the resilient modulus based
only on the subgrade classification is not completely reliable.
One can also observe that the subgrade resilient moduli obtained
with the iterative procedure are usually higher than the values recom-
mended in the design guide. The reason for this discrepancy is that the
stress state at 18 in. below the surface is too low and does not nec-
essarily adequately represent the stress state in the system. The resilient
moduli obtained for the subbase are significantly lower because the
stress state is calculated 2 in. below the bottom of the base layer.
Determination of the optimum location for the stress state calcu-
lation is beyond the scope of this paper. These recommendations
should be developed by comparison of pavement distresses (cracking
and rutting) predicted with the responses computed with the nonlinear
axisymmetric program (Level 1 input) and those with the MLET

TABLE 2 Subgrade and Subbase Resilient Moduli

Subbase MR, psi Subgrade MR, psi

2002 DG
Location AASHTO Class Recommended Ranges Mean Value Ranges Mean Value Ranges

City of Rochester A-7-6 5,000–13,500 13,113 10,0001–17,696 18,677 15,647–21,314

City of Rochester A-6 13,500–24,000 11,167 10,000–12,977 13,851 12,516–14,976
Isanti County A-6 13,500–24,000 20,892 10,000–30,0002 29,659 26,540–30,000
Isanti County A-6 13,500–24,000 18,602 10,000–26,092 28,007 23,595–30,000
Isanti County A-6 13,500–24,000 14,338 10,000–17,250 18,574 16,775–20,131
Itasca County A-6 13,500–24,000 17,263 10,000–23,475 24,018 20,804–26,734
Kandiyohi County A-6 13,500–24,000 15,949 10,000–20,253 22,333 19,490–24,889
Kandiyohi County A-6 13,500–24,000 15,428 10,000–19,564 21,145 18,596–23,377
Lyon County A-6 13,500–24,000 11,790 10,000–15,247 16,596 13,791–19,077
Lyon County A-6 13,500–24,000 21,578 10,000–30,000 29,664 26,872–30,000
Medford Roundabout A-6 13,500–24,000 13,547 10,000–16,534 18,428 16,301–20,317
MnRoad Cell 33 A-6 13,500–24,000 19,464 10,000–30,000 29,271 24,957–30,000
MnRoad Cell 34 A-6 13,500–24,000 13,536 10,000–16,715 17,751 15,862–19,346
MnRoad Cell 35 A-6 13,500–24,000 13,612 10,000–18,492 20,030 16,656–23,036
St. Louis County A-7-6 5,000–13,500 14,938 10,000–19,983 19,745 17,380–21,618
St. Louis County A-7-5 8,000–17,500 15,268 10,000–18,733 18,547 17,112–19,624
Stearns County A-6 13,500–24,000 10,000 10,000–10,000 8,274 7,245–9,212
Stearns County A-6 13,500–24,000 12,762 10,000–17,945 20,035 15,652–24,142
Wright County A-7-6 5,000–13,500 10,327 10,000–11,759 12,619 10,382–14,583
Wright County A-6 13,500–24,000 10,000 10,000–10,000 7,887 6,535–9,104
Wright County A-6 13,500–24,000 10,000 10,000–10,000 5,322 4,500–6,081
Wright County A-7-6 5,000–13,500 13,359 10,000–17,808 19,136 16,128–21,782
Wright County A-7-6 5,000–13,500 13,113 10,000–17,696 26,328 24,426–27,868

1
Minimum value for subbase MR
2
Maximum value for subgrade and subbase MR
Khazanovich, Celauro, Chadbourn, Zollars, and Dai 165

program (Level 2 input). However, the foregoing discussion clearly
illustrates the importance of defining this location. The current version
of the MEPDG lacks specific recommendations for this matter.
EFFECT OF PAVEMENT STRUCTURE
ON RESILIENT MODULUS SELECTION
As was stated earlier, the same k1-k2-k3 model can result in different
resilient moduli. The stiffness of the structure above the subgrade
(thickness and stiffness of the AC and base layers) may have a
significant effect on the predicted moduli of unbound materials. To
evaluate this effect, the resilient moduli were compared with the
effective flexural stiffness of the pavement system defined as follows:
Deff = EAC i hAC
3
+ Ebase i hbase
3
(11)
be attributed to change in the effective flexural stiffness. Therefore,
higher subgrade moduli may be used as the Level 2 input for design
of thicker and stiffer AC pavements. However, since the resilient
modulus–flexural stiffness relationship depends on the subgrade
material properties (parameters k1, k2, and k3), resilient modulus testing
is required to justify the increase in subgrade resilient modulus.
In Minnesota an AC modulus of elasticity varies significantly
during the year. Therefore, an effective flexural stiffness may vary
from month to month. A significant effect of the effective flexural
stiffness on the subgrade resilient modulus also means that seasonal
adjustment of the subgrade modulus should account not only for
seasonal variations in moisture conditions but also for seasonal vari-
ations in AC modulus. The default seasonal adjustment procedure
does not account for this effect.

where CONCLUSIONS

Deff = effective pavement stiffness, The characterization of unbound materials in the MEPDG was
EAC = elastic modulus of asphalt layer, reviewed and applied to Minnesota subgrades. The main emphasis
hAC = thickness of asphalt layer, was on the collection of parameters k1, k2, and k3 for Minnesota fine-
Ebase = resilient modulus of base layer, and grained soil and the procedure for interpretation of the resilient mod-
hbase = thickness of base layer. ulus test to provide an input to the MLET analysis (Level 2 input). This
Figure 7 presents the resilient moduli versus the effective flexural is an important adaptation of the MEPDG because the guide rec-
stiffness for one of the Minnesota subgrades. It was found that the ommends measurement of resilient moduli from the laboratory test-
effect of the pavement structure on the resilient modulus may be ing, but the procedure of interpreting the test data to obtain an input
described with a high level of confidence by using the following for an MLET analysis lacks specifics.
simple predictive model: The procedure to convert the k1-k2-k3 model into a modulus of
elasticity for the Level 2 input was refined. Rapid solutions were
Mr = A i LnDeff
13
+B (12) developed to eliminate direct use of the MLET program and make
execution of the procedure less tedious. This development allowed
where A and B are fitting parameters. conducting a comprehensive sensitivity analysis of the effect of base,
Table 3 presents values of the regression parameters and a good- subbase, and subgrade k1-k2-k3 parameters on the resulting moduli
ness-of-fit characteristic for the subgrade materials. A very high R2 of elasticity. It was found that for the same pavement structure
(88% or more) was observed for 19 out of 23 subgrades. This find- and asphalt modulus, the subgrade modulus is mostly affected by
ing means that for the majority of the tested Minnesota subgrades, the subgrade k1-parameter, followed by the subgrade k3-parameter.
more than 88% of the variation in the predicted resilient moduli may Although base and subbase material properties affected the subgrade

25,000
y = 2762 Ln(Deff1/3) + 1011.5
R2 = 0.9598
20,000
Resilient Modulus, psi

15,000

10,000

5,000

0
0 200 400 600 800 1,000 1,200 1,400 1,600
Effective Pavement Stiffness1/3, (psi*in)1/3

FIGURE 7 Resilient moduli versus effective flexural stiffness for A-7-6 subgrade in Rochester.
 166 Transportation Research Record 1947

TABLE 3 Parameters for the Resilient Moduli Models

Subgrade Number Location A B R2

1 City of Rochester 2,761.965 1,011.461 0.959799

2 City of Rochester 1,162.421 6,417.754 0.931872
3 Isanti County 982.33 23,371.95 0.402854
4 Isanti County 3,302.7 6,870.348 0.843207
5 Isanti County 1,573.928 8,504.884 0.921337
6 Itasca County 2,896.655 5,482.509 0.964598
7 Kandiyohi County 2,539.942 6,080.855 0.923769
8 Kandiyohi County 2,268.813 6,628.459 0.93482
9 Lyon County 2,560.032 223.0965 0.953461
10 Lyon County 938.4947 23,656.56 0.428909
11 Medford Roundabout 1,871.449 6,456.11 0.913378
12 MnRoad Cell 33 1,819.023 17,630.19 0.542029
13 MnRoad Cell 34 1,662.978 7,113.909 0.943121
14 MnRoad Cell 35 3,081.767 317.0227 0.952096
15 St. Louis County 2,123.267 6,161.618 0.977855
16 St. Louis County 1,248.64 10,558.66 0.97669
17 Stearns County 897.2161 2,538.019 0.88286
18 Stearns County 4,114.52 −6,282.26 0.953069
19 Wright County 2,060.483 −556.573 0.96136
20 Wright County 1,231.764 12.9514 0.943927
21 Wright County 727.4203 673.1695 0.899084
22 Wright County 2,727.254 1,691.147 0.951131
23 Wright County 1,672.368 15,621.62 0.963422

modulus of elasticity, the effect was less significant than the effect
of these two parameters.
Analysis of the test results for 23 samples collected from several
Minnesota locations indicated that the k1-k2-k3 parameters may vary
in wide ranges. These 23 combinations of the parameters were
converted into subgrade moduli of elasticity for a wide range of
pavement structures. The obtained elastic moduli were compared
with the MEPDG recommended ranges for subgrade modulus of
elasticity based on the soil classification (Level 3 inputs). The MEPDG
Level 3 ranges—recommended moduli of elasticity based on the
soil classification—were found to be reasonable.
It was found that the effective pavement stiffness may significantly
affect the subgrade resilient modulus. Since the AC modulus may
vary significantly, especially in a climate such as that of Minnesota,
it is important to adjust the subgrade modulus not only for the change
in moisture conditions but also for the change in the effective pave-
ment stiffness. Currently, the default seasonal adjustment procedure
accounts only for seasonal variation in moisture conditions.
REFERENCES
1. Part 2: Design Inputs. In 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. http://trb.org/mepdg/guide.htm.
2. Part 3: Design Analysis. In 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. http://trb.org/mepdg/guide.htm.
3. NCHRP Research Results Digest 285: Laboratory Determination of
Resilient Modulus for Flexible Pavement Design. TRB, National Research
Council, Washington, D.C., 2004. http://www4.trb.org/trb/crp.nsf/All+
Projects/NCHRP+1-28A.
4. Desai, C. S. User’s Manual for the DSC-2D Code for the 2002 Design
Guide. Arizona State University, 2000.
5. Uzan, J. JULEA Linear Elastic Analysis Computer Program. U.S. Army
Waterways Experiment Station, Vicksburg, Miss., 1989.
6. Larson, G., and B. J. Dempsey. Integrated Climatic Model: Version 2.0.
Report DTFA MN/DOT 72114. University of Illinois, Urbana–Champaign,
and Newmark Civil Engineering Laboratory, 1997.
7. Witczak, M. W., D. Andrei, and W. N. Houston. Resilient Modulus
as Function of Soil Moisture: Summary of Predictive Models. NCHRP
Project 1-37 A, Inter Team Technical Report (Seasonal 1). TRB, National
Research Council, Washington, D.C., 2000.
8. Witczak, M. W., W. N. Houston, C. E. Zapata, C. Richter, G. Larson, and
K. Walsh. Improvement of the Integrated Climatic Model for Moisture
Content Predictions. NCHRP Project 1-37 A Inter Team Technical Report
(Seasonal 4). TRB, National Research Council, Washington, D.C.,
June 2000.
9. Smith, B. E., and M. W. Witczak. Prediction of Equivalent Granular Base
Moduli Incorporating Stress Dependent Behavior in Flexible Pavement.
Journal of Transportation Engineering, ASCE, Vol. 107, No. 6, 1981,
pp. 635–652.
The Flexible Pavement Design Committee sponsored publication of this paper.

View publication stats