You are on page 1of 9

18 Paper No.

970539 TRANSPORTATION RESEARCH RECORD 1577

Anisotropic Modeling of Granular


Bases in Flexible Pavements
EROL TUTUMLUER AND MARSHALL R. THOMPSON

A new cross-anisotropic model is proposed to predict the performance define mainly the vertical load-induced anisotropic behavior under
of granular bases in flexible pavements. A cross-anisotropic represen- axisymmetric conditions.
tation has different material properties (i.e., elastic modulus and Finite element solutions of a conventional flexible pavement
Poisson’s ratio) assigned in the horizontal and vertical directions.
obtained from the recently developed GT-PAVE program (4) resulted
Repeated-load triaxial tests with vertical and lateral deformation mea-
surements can be used to establish these anisotropic properties. Sim- in very low to vanishing predicted radial tensile stresses in the gran-
ple stress-dependent granular material models, obtained from analysis ular base when modeled using the nonlinear cross-anisotropic
of the laboratory test data, are used in a nonlinear finite element pro- approach. The magnitudes of both the horizontal and shear stiffnesses
gram, named GT-PAVE, to predict pavement responses. The horizon- throughout the base were found to be only small fractions of the
tal and shear stiffnesses are typically found to be less than the vertical. vertical stiffness, with horizontal stiffness being mainly the lowest
The nonlinear anisotropic approach is shown to account effectively for
under the wheel load. The effects of compaction-induced residual
the dilative behavior observed under the wheel load and the effects of
compaction-induced residual stresses. The main advantage of using a stresses locked in granular bases were also of significance, especially
cross-anisotropic model in the base is the drastic reduction or elimi- when calculating horizontal stiffnesses. Such stresses offset any
nation of significant tensile stresses generally predicted by isotropic low-magnitude tensile stresses and provided adequate confinement
linear elastic layered programs. radially away from the wheel load.

An aggregate base course in a flexible pavement provides load dis-


tribution through aggregate interlock. The load transfer is achieved MATERIAL CHARACTERIZATION
through compression and shear forces between the particles oriented
in a chainlike manner. Because the tensile forces cannot be trans- The resilient modulus (MR) in unbound aggregate base is well-
ferred from grain to grain, when such forces act in the horizontal known to be dependent on the stress state to which each small ele-
direction, the behavior of the base course is significantly affected by ment of material is subjected (6,9). Since the stress states vary both
a directional dependency of material stiffnesses that can be accom- vertically and radially within a layer, the linear elastic assumption
modated by using an anisotropic approach. A better basic under- of assigning a single modulus to the entire layer does not correctly
standing of this kind of anisotropic behavior of unbound aggregate model layer stiffnesses. Therefore, in most cases, ELPs predict sig-
bases and subbases helps to define more accurately the stress, strain, nificant tensile stresses in the unbound aggregate base. This is in
and deformation responses of flexible pavements. contradiction with very little tension-taking capability of granular
At the present time, mechanistic pavement design methods—such materials.
as the Asphalt Institute (1), the Shell Method (2), and the LEDFAA Stiffness stress dependence can be reasonably represented as
(3)—use linear elastic programs (ELPs) to predict wheel-load- power functions of the stress states in the granular base. The K-θ
induced deflections, stresses, and strains in the pavement. The linear model (10), which relates resilient modulus with bulk stress only,
elastic methods consider neither the horizontal variation of material has been the most frequently used stress-dependent granular mate-
stiffnesses nor the effects of compaction-induced residual stresses rial characterization model in the United States. The K-θ model,
throughout the base course. The nonlinear stress-dependent behav- however, can give inaccurate results since it neglects the important
ior of the unbound aggregate pavement layers has been well docu- effect of shear stress on the resilient modulus (11). The Uzan model,
mented (4–6). Furthermore, none of the linear elastic approaches on the other hand, considers shear stress effects in the granular
currently used considers the anisotropic behavior observed in gran- materials as shown here (11):
ular materials mainly due to compaction and the applied vertical KB KC
loading (7,8). θ
MR = K A    σd  (1)
A new, improved way of modeling granular bases using cross-  P0   P0 
anisotropic elasticity is described in this paper. Cross-anisotropy
where
enables a realistic assignment of in-plane and out-of-plane stiffnesses
in unbound aggregates. Repeated-load triaxial test results, which θ = σ1 + σ2 + σ3 = bulk stress,
included lateral deformation measurements performed on a variety σd = σ1 − σ3 = deviator stress,
of aggregate types, were obtained from the literature. From these P0 = unit reference pressure (1 kPa or 1 psi), and
results, nonlinear stress-dependent material characterization models KA, KB, KC = material constants obtained from repeated-load
were developed for the horizontal, vertical, and shear stiffnesses to triaxial tests performed on granular materials.

Department of Civil Engineering, University of Illinois at Urbana-Champaign, Because of its simplicity, the constants used in this model can be
205 North Mathews Avenue, Urbana, Ill. 61801. readily determined from resilient modulus tests such as AASHTO
Tutumluer and Thompson Paper No. 970539 19

T294-94. When compared with the more complicated shear and rate testing. In 1995 Tutumluer and Barksdale (9) modeled the same
volumetric stress-strain contour model (5), this type of model gives test sections by employing cross-anisotropic resilient properties in
reasonably good agreement and therefore is recommended for the base layer and using an Uzan (Equation 2) type nonlinear
use as a practical model to characterize granular bases in flexible response model. Considerably lower horizontal tensile stresses were
pavements (4,11). predicted in the granular base when the horizontal resilient modulus
was equal to 15 percent of the vertical resilient modulus. Using this
anisotropic modeling approach, reasonably good agreement was
CROSS-ANISOTROPY achieved with measured values of the resilient behavior for up to
eight response variables at the same time.
The behavior of a granular medium depends at any point on the Lo and Lee (18) studied experimentally the response of a gran-
arrangement of particles, which is usually determined by aggregate ular soil tested along constant stress increment ratio paths under
characteristics, construction methods, and loading conditions. For axisymmetric conditions. Deformation behavior for unloading was
granular bases in flexible pavements, an apparent anisotropy is found to be essentially elastic in nature. The elastic response was
induced in the fabric during construction by aggregate placement anisotropic and the degree of anisotropy increased with the prin-
and then loading from the compaction equipment. The granular cipal stress ratio. Based on these findings, a cross-anisotropic
layer, therefore, becomes stiffer in the vertical direction than in the model, with the degree of anisotropy increasing with the principal
horizontal direction even before the wheel load on the pavement stress ratio, was proposed to model elastic granular material
imposes further anisotropic loading. behavior.
Several researchers reported the effects of anisotropic behavior of Karasahin et al. (19) also reported results of a study in which the
cohesionless soils on the computed stress-strain response. Borow- applicability of various resilient constitutive models of granular
icka (12) indicated an increase in the calculated vertical stresses near material was investigated for use in unbound base layers. An
the load when overburden stresses were considered to cause an ini- anisotropic volumetric-deviatoric type model by Elhannani (20) was
tial anisotropic material behavior. Similar results were obtained by found to give the best results for modeling the resilient behavior for
Barden (13) and Gerrard and Mulholland (14) when anisotropy was the following two loading conditions: (a) only the deviator stress
taken into account. was cycled, and (b) both deviatoric and confining pressures were
Using a highly sophisticated, true triaxial testing device, Desai cycled in a triaxial test.
et al. (7 ) performed extensive tests on three uniformly sized aggre-
gates used as ballast in track support structures. In each test, the mate-
rial was spooned into the cubical mold (10 × 10 × 10 cm) and then ANISOTROPIC PROPERTIES FROM
compacted by vibration. An apparent deviation from isotropy exhib- TRIAXIAL TESTS
ited by the test specimens was attributed to both material anisotropy
and specimen preparation, with the lowest strains measured in the The repeated-load triaxial compression test is currently the most
vertical direction of compaction. As the aggregate size was reduced commonly used method to measure the resilient and permanent
from 16 to 3 mm (0.63 in. to 0.12 in.), similar strain responses were deformation characteristics of unbound aggregates for use in pave-
observed in the two horizontal directions. This kind of behavior can ment design. The resilient modulus test is performed on a cylindrical
be modeled by using cross-anisotropy under axial symmetry. specimen of granular materials subjected to repeated axial compres-
An isotropic model has the same resilient material properties in sive (deviator) stresses. One of the main advantages of the triaxial
all directions. A cross-anisotropic representation, however, has test is that the axial and radial (or volumetric and shear) strains can
different resilient properties (i.e., resilient modulus and resilient be determined relatively easily. The standard AASHTO test proce-
Poisson’s ratio) in the horizontal and vertical directions. Five dure (AASHTO T294-94; “Resilient Modulus Testing of Unbound
cross-anisotropic material properties are needed to define an Granular Base/Subbase Materials and Subgrade Soils”—SHRP Pro-
anisotropic material under conditions of axial symmetry. A gen- tocol P46) provides specifications for specimen axial deformation
eral formulation of a cross-anisotropic layered system in terms of measurements using externally mounted linear variable differential
the in-plane resilient modulus (M Rr ), normal to layered strata transformers (LVDTs). Lateral deformation measurement is not
resilient modulus (M R) and Poisson’s ratios (νr and νz) has been
z
included in AASHTO T294-94.
given by Zienkiewicz and Taylor (15). The variables n and m are Determination of lateral strains in a triaxial specimen is essential
commonly substituted for horizontal modulus (M Rr ) and shear for characterizing the anisotropic elastic properties of granular bases.
z
modulus (G R ) in the formulation; they represent the ratios of The specimen lateral deformation measurements require certain
horizontal modu-lus to vertical modulus and shear modulus to experimental setup. These measurements were taken in the past by
vertical modulus, respectively. using either Bison coils attached to the specimen (21,22) or radial
In 1977 Lo et al. (16) performed triaxial tests on undistributed strain gauges attached to the specimen at midheight (23). Anisotropic
Leda clay with the specimens oriented in 0, 45, and 90 degrees with resilient response can be defined from triaxial test data with measured
the horizontal. The typical values of the five elastic anisotropic con- vertical and lateral deformations as follows:
stants were defined for the first time from locally measured strains.
The results obtained by Lo et al. (16 ) suggested that the stiffness on Vertical resilient modulus:
a vertical plane in the Leda clay was appreciably higher than on a
horizontal plane. MRz = σ d /e axial (standard definition ) (2)
Barksdale et al. (17 ) observed from instrumented test sections that
a linear cross-anisotropic model of an unbound aggregate base is at Horizontal resilient modulus:
least equal to and perhaps, for predicting general pavement response,
better than the simplified contour model (5), which requires elabo- MRr = σ 3 /e lateral (3)
20 Paper No. 970539 TRANSPORTATION RESEARCH RECORD 1577

Resilient shear modulus:

GRz = σ d / 2 (e axial − e lateral ) ( 4)

where the horizontal resilient modulus (M Rr ) is a new definition.


In addition to the moduli values, the two Poisson’s ratios, νr and
νz, have to be selected for the elastic anisotropic analysis of granu-
lar material behavior. These constants νr and νz are defined as the
Poisson’s ratios for strain in any horizontal direction due to a hori-
zontal stress at right angles, and for strain in the vertical direction
due to a horizontal direct stress, respectively. Using the quadratic
form for the matrix of elastic cross-anisotropic constants, Pickering
(24) deduced the bounding values of the Poisson’s ratios necessary
to ensure that the resulting strain energy was positive.

ANISOTROPIC RESILIENT MODEL


DEVELOPMENT

The Uzan type equation (Equation 1) can be used to express the


stress dependency of the vertical resilient modulus (Equation 2)
within the context of cross-anisotropy. In addition, an anisotropic
representation of the granular bases requires that the resilient hori-
zontal and shear stiffnesses (Equations 3 and 4) now also be mod-
eled using lateral deformation measurements obtained from the
triaxial tests. As expected, stress-dependent behavior is observed
in the horizontal and shear planes, and the behavior can also be
modeled as nonlinear power functions of stress states.
Repeated-load triaxial test results including lateral strain data
from the works of Hicks (23), Allen (25), and Crockford et al. (26)
have been recovered and used in this study. A total of 50 individual
tests were performed in these studies on gravels, partially crushed,
and crushed aggregates at varying densities and saturation levels.
The analysis using all the data sets consisted of two stages. In the
first stage, the resilient horizontal, vertical, and shear moduli
obtained by using Equations 2 to 4 were plotted with various tri- FIGURE 1 Variation of measured stiffnesses with strain levels
for crushed stone HD1 (1 in. = 25.4 mm; 1 psi = 6.89 kPa).
axial test variables such as the specimen strain response and the
stress conditions (i.e., deviator and bulk stresses).
Typical variations of the three moduli with strain levels are pre- APPLICATION OF MODEL IN GRANULAR BASES
sented in Figure 1 for the high-density, high-quality partially satu-
rated crushed stone (HD1) tested by Allen (25). As the shear modulus The recently developed nonlinear finite element program GT-PAVE
increases with increasing strains, which is similar to the generally (4) was used to employ the cross-anisotropic model in granular bases
observed trends for the vertical modulus at different confining pres- of conventional flexible pavements. The GT-PAVE program runs on
sures, the horizontal modulus tends to decrease with increasing radial a PC, considers the cross-anisotropic behavior exhibited by unbound
strains. This kind of variation with strain levels can also be perceived aggregate bases/subbases, and also handles residual compaction
for the moduli when plotted with increasing bulk and deviator stresses. Isoparametric eight-node quadrilateral elements are used in
stresses. The observed horizontal modulus reduction in the large the program to analyze a flexible pavement consisting of either
strain zone is in accordance with the dilation phenomenon often linear or nonlinear elastic layers. The details on the nonlinear solu-
encountered in the granular bases under the wheel load (22,27). tion technique and the base and subgrade characterization models
In the second stage of data analysis, several combinations of the used in the GT-PAVE program were described elsewhere (4,9).
stress variables were investigated for modeling the stiffnesses as The conventional pavement section modeled here was asphalt
functions of stress states. Finally, a Uzan type model, which relates concrete (AC) surfacing 89 mm (3.5 in.) thick underlain by a 203-mm
the moduli to both bulk and deviator stresses, was found to give a (8.0-in) unstabilized aggregate base layer placed over a silty sand
good statistical fit with high correlation coefficients (R2 = 0.95 to
0.99, n = 19) when the model parameters KA, KB, and KC were
TABLE 1 Model Parameters for Anisotropic Moduli
obtained from the multiple regression analysis of the individual test
data. The three moduli, each modeled by the same stress-dependent Modulus Coefficient θ Exponent σd Exponent
functional form, therefore have the model parameters given in
Table 1, where K1 to K9 replace for each modulus the model param- M Rr K1 = constant K2 K3
eters KA, KB, and KC given in Equation 2. The stiffness ratios m and M Rz K4 = constant K5 K6
n can be expressed in the same form as presented in Table 2. G Rz K7 = constant K8 K9
Tutumluer and Thompson Paper No. 970539 21

TABLE 2 Model Parameters for Horizontal and layers were later compared with the limiting Mohr-Coulomb
Shear Stiffness Ratios values not to exceed shear strength of the materials.
Ratio Coefficient θ Exponent σd Exponent

n = M Rr /M Rz K1/K4 = constant K2–K5 K3–K6 INTERPRETATION OF RESULTS


m = G Rz /M R K7/K4 = constant
z
K8–K5 K9–K6
The results of the present analysis using the GT-PAVE program are
of significance since three constitutive models were used for the first
subgrade 1270 mm (50 in.) thick. The geometry and the material time in a nonlinear cross-anisotropic base to completely define the
properties used were identical to those of the full-scale conventional resilient granular material behavior in vertical, horizontal, and shear
test sections tested at the Georgia Tech test pit facility (28,29). The planes. Unlike isotropic analysis, the horizontal stiffness can now
wheel load was applied as a uniform pressure of 689 kPa (100 psi) take values different than the vertical. These stiffnesses are not
over a circular area of radius 116 mm (4.55 in.). assumed in the base layer, but predicted by Uzan type models
A 140-element, 475-node axisymmetric finite element mesh was obtained directly from the triaxial specimen behavior.
used to analyze the section as a nonlinear elastic layered system. The Figure 2 shows a contour plot of the final vertical resilient mod-
subgrade and the unbound aggregate base were treated as nonlinear ulus distribution achieved at the end of nonlinear analysis in the
elastic materials, while the AC surfacing was modeled as linear elas- top portion of the conventional section. A decrease in modulus
tic. The AC layer was assigned an elastic modulus of 1724 MPa from 206.9 MPa (30 ksi) at the top of the centerline base down to
(250 ksi) and a Poisson’s ratio of 0.35. The bilinear model (6) was 96.5 MPa (14 ksi) at the bottom is a result of the stress-dependent
used for the subgrade with the following parameters: (a) breakpoint stiffness models used in the analysis. A similar reduction in aggre-
deviator stress of 17 kPa (2.5 psi) corresponding to a modulus of gate base stiffness is also observed in the horizontal direction.
40.7 MPa (5.9 ksi), and (b) slopes on the left and right sides of the The subgrade stiffness at the top was calculated to be about
breakpoint being 11 640 kPa/kPa and 26.7 kPa/kPa, respectively. 41.4 MPa (6 ksi).
The 203-mm (8-in.) unstabilized aggregate base, which is of main To compare radial stresses predicted by the nonlinear cross-
concern in this paper, was assigned a complete set of cross- anisotropic base model with the stresses commonly computed from
anisotropic Uzan type model parameters as obtained from triaxial the ELPs, an isotropic linear elastic analysis was performed using
test results for the high-density, high-quality crushed stone (HD1). an average constant modulus of 151.7 MPa (22 ksi) from nonlinear
The following anisotropic parameters were used for HD1: analysis in the base layer. The variations of radial stresses through
the 203-mm (8-in.) base course obtained from both analyses are
MRr : shown in Figure 3. High centerline tensile stresses of up to 124 kPa
(18 psi) calculated at the bottom of the base from the linear
K1 = 40.76 MPa (or 287 psi ), K2 = 3.416, isotropic run were reduced to 10 kPa (1.5 psi) using the nonlinear
K3 = − 2.808 (n = 19, R2 = 0.992) (5)
cross-anisotropic approach. Similarly, at one load radius [116 mm
(4.55 in.)] radially away from centerline, the tensile stresses pre-
dicted by the nonlinear anisotropic model almost vanished, taking
MRz :
values of about 3.5 kPa (0.5 psi) throughout the depth. The linear
K4 = 533.92 MPa (or 2318 psi ), K5 = 0.640, elastic isotropic solution, however, indicated tensile stresses as
high as 90 kPa (13 psi) at the same radial offset.
K6 = 0.065 (n = 19, R2 = 0.995) (6) This drastic reduction in the predicted horizontal tension could be
achieved only when using nonlinear cross-anisotropy as compared
GRz : with an isotropic analysis. The low predicted tensile stresses in the
K7 = 143.75 MPa (or 754 psi ), K8 = 0.834, range of 7 to 14 kPa (1 to 2 psi) can be accommodated by either
(a) the counteracting effects of the compressive residual stresses
K9 = − 0.167 (n = 19, R2 = 0.996) ( 7) (which were not included at this time in the analysis) locked in the
base layer due to compaction (11,30), or (b) the limited tensile
The analyses performed using these models were mainly carried strength capability of granular materials under confinement (4,31).
out with U.S. customary units. The constant Poisson’s ratios used Figure 4 illustrates throughout the base the distributions of the
were νr = 0.3 and νz = 0.1, which satisfied the requirements of the predicted horizontal and shear stiffness ratios n and m, respectively.
positive strain energy (24). A wide range and combination of values The ratios, shown as contour plots, were obtained from the final
considered for Poisson’s ratios proved to have negligible effect on stress state using the nonlinear models with the previously given
the computed granular material stiffnesses. parameters for HD1 crushed stone. Both ratios slightly increase
The nonlinear analysis using the GT-PAVE program includes radially away from the centerline, with horizontal stiffness ratio
a combined incremental and iterative procedure in which the being in the range of 9 to 15 percent and the shear stiffness ratio
wheel load was applied in 10 increments. For each increment, between 29 to 34 percent of the vertical modulus. These ratios com-
iterations were performed on the vertical modulus, which was puted by GT-PAVE using the anisotropic models are indeed in very
checked each time for convergence criteria in the granular base good agreement with the ones formerly assigned in the aggregate
layer. The three moduli were computed from the stress state base for modeling the resilient behavior of the Georgia Tech full-
obtained in the previous iteration using the cross-anisotropic scale conventional test sections (4,9). In that study, reasonably good
model. The final resilient displacements, strains, and stresses predictions of up to eight response variables were achieved at the
were achieved in the section when the iterations converged for the same time by using a constant 15 percent for the horizontal stiffness
total load. The shear stresses obtained in the base and subgrade ratio and a constant 35 percent for the shear stiffness ratio.
FIGURE 2 Vertical modulus distribution throughout crushed stone (HD1) base
(1 in. = 25.4 mm; 1 psi = 6.89 kPa).

FIGURE 3 Variation of predicted radial stresses through 203-mm


(8 in.) base (1 in. = 25.4 mm; 1 psi = 6.89 kPa).
Tutumluer and Thompson Paper No. 970539 23

FIGURE 4 Stiffness ratio distributions throughout crushed stone (HD1)


base: top, horizontal stiffness ratio; bottom, shear stiffness ratio (1 in. = 25.4
mm; 1 psi = 6.89 kPa).

EFFECTS OF RESIDUAL STRESSES The radial stress distribution with depth in base predicted in the
presence of the 21-kPa (3-psi) residual stresses is also shown in
The influence of the compaction-induced residual stresses on the pre- Figure 3. The unbound aggregate base is completely in compression
dicted horizontal tensile stresses in granular bases has been investi- with radial compressive stresses increasing away from the center-
gated by several researchers (4,11,30). Selig (30) proposed one line. The total elimination of tensile stresses is achieved by employ-
possible explanation to the “no tension” problem, stating that the ing the nonlinear anisotropy in the granular base and considering the
existence of high horizontal compressive residual stresses in a base offsetting effects of the compaction-induced residual stresses.
layer offsets the incremental tensile stresses predicted by the elastic The major effect of the additional 21-kPa (3-psi) residual stress
solution. The magnitudes of these horizontal residual stresses were was observed to be on the predicted horizontal stiffnesses. Figure 5
recently measured in the field as high as 21 kPa (3 psi) in the shows the predicted horizontal-to-vertical modular ratio distribution
unbound aggregate due to 89-kN (10-ton) vibratory compaction (32). throughout the aggregate base in the presence of the 21-kPa residual
To demonstrate the offsetting effects of the residual stresses on stresses. Compared with the distribution indicated in Figure 4 (top),
the horizontal tension in granular bases, the conventional flexible the ratios are still low under the wheel load (dilative behavior), vary-
pavement problem of the previous section was reanalyzed using the ing between 0.08 to 0.12 with depth at the centerline. However, at
GT-PAVE program. Everything else being the same, a constant about two load radii away from the centerline, they increase rapidly,
compressive horizontal residual stress of 21 kPa (3 psi) was exceeding the previously maximum 0.15 value. Because of the pres-
assumed to exist initially throughout the granular base before the ence of the residual stresses, confining pressures are in fact higher
wheel load was applied. The analysis performed resulted in slightly radially away from the centerline, which results in minimal lateral
higher vertical stiffnesses in the base layer with a centerline modu- strains. As a result, horizontal stiffnesses obtained from the model
lus distribution almost identical to the one shown in Figure 2. Sim- rapidly increase at low lateral strains (see Figure 2), thus causing high
ilarly, the shear stiffnesses ranged from 28 to 34 percent of the modular ratios. To realistically model the stiffnesses essentially away
vertical stiffnesses. The vertical stress on subgrade was about 14 kPa from the wheel load, the modular ratio was not allowed to exceed 1
(2 psi) lower than the previous 97 kPa (14 psi) value, thus causing a in the layer even though the stress-dependent horizontal modulus
reduction in the subgrade deflection. model may actually predict values higher than the vertical moduli.
24 Paper No. 970539 TRANSPORTATION RESEARCH RECORD 1577

FIGURE 5 Horizontal stiffness ratio distribution throughout crushed stone (HD1)


base in presence of 21-kPa (3-psi) compressive residual stresses (1 in. = 25.4 mm;
1 psi = 6.89 kPa).

TYPICAL VARIATIONS OF STIFFNESS RATIOS

To obtain a better understanding of the typical variations of hori-


zontal and shear stiffness ratios, anisotropic material models, simi-
lar to those given in Equations 5 to 7 for HD1, were developed for
the other 49 triaxial test results from Hicks (23), Allen (25), and
Crockford et al. (26). It was generally observed that Uzan type
stress-dependent models, when used for modeling the horizontal
and shear stiffness ratios, resulted in a constant term (K1/K4 or K7/K4)
almost equal to the average ratios predicted by the finite element
analysis throughout the base. This observation was indeed sup-
ported by the fact that the exponents of two stresses (i.e., bulk and
deviator stresses) in the models were taking values very close to
each other in magnitude but with opposite signs, thus somewhat
canceling the effects of each other at similar stress levels.
Figure 6 shows for the 50 test results the deviator stress exponents
(K3-K6 or K9-K6) plotted with the bulk stress exponents (K2-K5 or
K8-K5) as obtained from the horizontal and shear stiffness ratio mod-
els. In both plots, the data points are generally centered on the equal-
ity line. They are similar in magnitude but opposite in sign. The
scatter of data is greater in Figure 6 (bottom) because of the nar-
rower range of values that the exponents takes. A common trend
also observed in both plots is that the partially to fully saturated
aggregates tend to have larger deviator exponents than the bulk
stress exponents. Overall, these plots indicate that the deviator and
bulk stress terms have less effect at similar bulk and deviator
stresses in determining the stiffness ratios when compared with the
governing role of the constant ratio term in the models.
The constant ratios K1/K4 and K7/K4 used in the stiffness ratio mod-
els are then plotted with the model constant K4 of the vertical resilient
modulus model in Figure 7. The constant ratio K1/K4 for the hori-
zontal stiffness [Figure 7 (top)] typically takes values between 0.03
and 0.21 for dry to partially saturated aggregates having low to high
densities. In the case of fully saturated aggregates, the ratio is
observed to be higher. The high ratio of 0.8 is attributed to an anom- FIGURE 6 Variation of stress exponents in stiffness ratio
aly in the test results. The constant ratio K7/K4 for the shear stiffness, models: top, horizontal stiffness ratio model; bottom, shear
on the other hand, [Figure 7 (bottom)] typically ranges from 0.18 to stiffness ratio model.
Tutumluer and Thompson Paper No. 970539 25

predicted tensile stresses in the unbound aggregate base when com-


pared with the high values commonly estimated by isotropic linear
elastic layered programs.
5. The dilative behavior under the wheel load was adequately
modeled in the granular base. Low horizontal stiffnesses were pre-
dicted in the analyses both in the presence and absence of com-
paction-induced residual stresses. When residual stresses are
included in the nonlinear anisotropic analysis, they offset the hori-
zontal tensile stresses and provide confinement at locations radially
away from the wheel load.
6. The analyses performed for a variety of aggregate types and
properties used in the granular layer typically resulted in horizontal
stiffnesses varying between 3 and 21 percent of the vertical and shear
stiffnesses between 18 and 35 percent of the vertical throughout
the base.

ACKNOWLEDGMENTS

This paper was prepared from a study conducted in the Center of


Excellence for Airport Pavement Research. Funding for the Center
of Excellence is provided in part by FAA. The Center of Excellence
is maintained at the University of Illinois at Urbana-Champaign,
which works in partnership with Northwestern University and FAA.
Patricia Watts is the FAA Program Manager for Air Transportation
Centers of Excellence, and Satish Agrawal is the FAA Technical
Director for the Pavement Center.

REFERENCES

1. Research and Development of the Asphalt Institute’s Thickness Design


Manual (MS-1), 9th ed. Research Report 82-2. Asphalt Institute, 1982.
FIGURE 7 Variation of constant ratios in stiffness ratio models 2. Claussen, A. I. M., J. M. Edwards, P. Sommer, and P. Ugo. Asphalt Pave-
ment Design—The Shell Method. Proc., 4th International Conference on
with K4: top, horizontal stiffness ratio model; bottom, shear
the Structural Design of Asphalt Pavements, Vol. 1, 1977, pp. 39–74.
stiffness ratio model.
3. Airport Pavement Design for the Boeing 777 Airplane. Advisory Circular
(AC) No. 150/5320-16. U.S. Department of Transportation, 1995.
4. Tutumluer, E. Predicting Behavior of Flexible Pavements with Gran-
0.35 for most of the dry to partially saturated aggregate types. For ular Bases. Ph.D. dissertation. School of Civil and Environmental
fully saturated aggregates, it approaches 0.42, higher than the typi- Engineering, Georgia Institute of Technology, Atlanta, Sept. 1995.
cal values. Lo et al. (16 ) also reported similar increases for Leda clay 5. Brown, S. F., and J. W. Pappin. Analysis of Pavements with Granular
Bases. In Transportation Research Record 810, TRB, National Research
in the horizontal and shear stiffness ratios with increasing saturation
Council, Washington, D.C., 1981, pp. 17–23.
levels when calculated from the measured strains in undrained 6. Thompson, M. R., and R. P. Elliott. ILLI-PAVE-Based Response Algo-
triaxial tests. rithms for Design of Conventional Flexible Pavements. In Transportation
Research Record 1043, TRB, National Research Council, Washington,
D.C., 1985, pp. 50–57.
7. Desai, C. S., H. J. Siriwardane, and R. Janardhanan. Interaction and
CONCLUSIONS Load Transfer in Track Support Structures, Part 2: Testing and Consti-
tutive Modeling of Materials and Interfaces. Final Report, Contract
1. Cross-anisotropy enables a realistic assignment of in-plane DOT-US-05-80013. Office of University Research, U.S. Department of
and out-of-plane stiffnesses in stratified granular bases, with the Transportation, Jan. 1983.
8. Chan, F., R. D. Barksdale, and S. F. Brown. Aggregate Base Re-
vertical plane being the primary loading direction. inforcement of Surfaced Pavements. In Geotextiles and Geomembranes,
2. Repeated-load triaxial test results with axial and lateral Vol. 8, Elsevier Applied Science, 1989, pp. 165–189.
deformation measurements can adequately define axial and lateral 9. Tutumluer, E., and R. D. Barksdale. Behavior of Pavements with
specimen behavior. Granular Bases—Prediction and Performance. Proc., UNBAR4 Sym-
3. Using simple stress-dependent models, both the horizontal and posium (A. R. Dawson and R. H. Jones, eds.), Nottingham, UK, July
1995, pp. 173–183.
vertical stiffnesses can be modeled together with the shear modulus 10. Hicks, R. G., and C. L. Monismith. Factors Influencing the Resilient
to account for the intralayer stiffness variations within the granular Response of Granular Materials. In Highway Research Record 345,
bases. TRB, National Research Council, Washington, D.C., 1971, pp. 15–31.
4. The recently developed GT-PAVE finite element program was 11. Uzan, J. Characterization of Granular Material. In Transportation
Research Record 1022, TRB, National Research Council, Washington,
used to analyze the performance of a granular base layer when mod- D.C., 1985, pp. 52–59.
eled by such a nonlinear cross-anisotropic approach. The lower hor- 12. Borowicka, H. Pressure Distribution in a Halfspace with a Linearly
izontal stiffnesses resulted in a drastic reduction or elimination of Varying Modulus of Elasticity. Ingenieur-Archiv, Vol. 14, No. 2, 1943.
26 Paper No. 970539 TRANSPORTATION RESEARCH RECORD 1577

13. Barden, L. Stresses and Displacements in a Cross-Anisotropic Soil. 24. Pickering, D. J. Anisotropic Elastic Parameters for Soil. Geotechnique,
Geotechnique, Sept. 1963. Vol. 20, No. 3, 1970, pp. 271–276.
14. Gerrard, C. M., and P. Mulholland. Stress-Strain and Displacement 25. Allen, J. J. The Effects of Non-constant Lateral Pressures on the Resilient
Distributions in Cross-Anisotropic and Two-Layer Isotropic Elastic Response of Granular Materials. Ph.D. dissertation. Department of Civil
Systems. Proc., 3rd Conference of Australian Road Research Board, Engineering, University of Illinois, Urbana, May 1973.
Part 2, 1966. 26. Crockford, W. W., L. J. Bendana., W. S. Yang, S. K. Rhee, and S. P.
15. Zienkiewicz, O. C., and R. L. Taylor. The Finite Element Method, Senadheera, Modeling Stress and Strain States in Pavement Structures
Vol. 1: Basic Formulation and Linear Problems, 4th ed. McGraw Hill Incorporating Thick Granular Layers. Final Report, Contract F08635-
Book Co. (UK) Limited, 1989. 87-C-0039. Texas Transportation Institute, College Station, April 1990.
16. Lo, K. Y., G. A. Leonards, and C. Yuen. Interpretation and Significance 27. Uzan, J. Resilient Characterization of Pavement Materials. International
of Anisotropic Deformation Behavior of Soft Clays. Publication 117. Journal for Numerical and Analytical Methods in Geomechanics,
Norwegian Geotechnical Institute, Oslo, 1977. Vol. 16, 1992, pp. 453–459.
17. Barksdale, R. D., S. F. Brown, and F. Chan. NCHRP Report 315: Poten- 28. Barksdale, R. D., and H. A. Todres. A Study of Factors Affecting Crushed
tial Benefits of Geosynthetics in Flexible Pavements. TRB, National Stone Base Performance. Final Report SCEGIT-82-109. Georgia
Research Council, Washington, D.C., 1989. Institute of Technology, Atlanta, 1983.
18. Lo, S.-C. R., and I. K. Lee. Response of Granular Soil Along Constant 29. Tutumluer, E., and R. D. Barksdale. Inverted Flexible Pavement
Stress Increment Ratio Path. Journal of Geotechnical Engineering, Response and Performance. In Transportation Research Record 1482,
Vol. 116, No. 3, March 1990, pp. 355–376. TRB, National Research Council, Washington, D.C., 1995, pp. 102–110.
19. Karasahin, M., A. R. Dawson, and J. T. Holden. Applicability of 30. Selig, E. T. Tensile Zone Effects on Performance of Layered Systems.
Resilient Constitutive Models of Granular Material for Unbound Base Geotechnique, Vol. 37, No. 3, 1987, pp. 247–254.
Layers. In Transportation Research Record 1406, TRB, National 31. Heukelom, W., and A. J. G. Klomp. Dynamic Testing as a Means of
Research Council, Washington, D.C., 1993, pp. 98–107. Controlling Pavements During and After Construction. Proc., 1st Inter-
20. Elhannani, M. Modelisation et Simulation Numerique des Chaussees national Conference on the Structural Design of Asphalt Pavements,
Souples. Ph.D. thesis, University of Nantes, France, 1991. 1962, pp. 667–685.
21. Hussaini, M. M., and F. C. Townsend. Investigation of K0 Testing 32. Barksdale, R. D., and J. L. Alba. Laboratory Determination of Resilient
in Cohesionless Soils. Technical Final Report S-75-16. U.S. Army Modulus for Flexible Pavement Design. Interim Report 2. NCHRP,
Engineer Waterways Experiment Station, Vicksburg, Miss., 1975. TRB, National Research Council, Washington, D.C., 1993.
22. Allen, J. J., and M. R. Thompson. Resilient Response of Granular
Materials Subjected to Time-Dependent Lateral Stresses. In Trans- The contents of this paper reflect the views of the authors, who are respon-
portation Research Record 510, TRB, National Research Council, sible for the facts and accuracy of the data presented within. The contents
Washington, D.C., 1974, pp. 1–13. do not necessarily reflect the official views and policies of FAA. This paper
23. Hicks, R. G. Factors Influencing the Resilient Properties of Granular does not constitute a standard, specification, or regulation.
Materials. Ph.D. dissertation. Institute of Transportation and Traffic
Engineering, University of California, Berkeley, May 1970. Publication of this paper sponsored by Committee on Mineral Aggregates.

You might also like