Anisotropic Behavior of Granular Materials: Impact on

Flexible Pavement Design

Abstract.: This study investigated the effect of anisotropic behavior of granular

material in flexible pavement and quantify the effect on the overall design of the
flexible pavement. The major scope of this study included design of flexible
pavement based on conventional design using layered-elastic theory and aniso-
tropic behavior of granular materials. These properties were considered through
resilient modulus of granular and modelled through finite element modeling.
Strain characteristics of pavement was simulated using ABAQUS and perfor-
mance was predicted based on IRC 37: 2018 pavement design guidelines. It was
found that anisotropic behavior of granular materials overestimated the material
properties which can potentially lead to premature failure.

Keywords: Pavement Design, Anisotropic Behavior, Finite Element Modeling,

Design Life.

1 Introduction

Flexible pavement consists of bituminous mixtures placed on compacted aggregates

over soil subgrades. The load-transfer mechanism of flexible pavement employs distri-
bution of load through different layers where the load magnitude diminishes as the load
is transmitted downwards. Using this framework, the mechanics characteristics of pave-
ments such as: stress and strain at different locations are estimated. In general, these
mechanics characteristics are used as inputs in mechanistic-empirical design of flexible
pavement.
Globally, flexible pavement is designed based on mechanistic pavement design
methodology. In this system, strain is calculated based on layered-elastic theory and
then the strains at different locations are considered as indicators of pavement perfor-
mances. For instance, in India, flexible pavement design is conducted based on majorly
two performance criteria. These performance criteria include: (a) rutting and (b) fatigue
cracking. Even though these two criteria depend on several other design variables such
as traffic level, reliability, percentage volume of air voids, and percentage volume of
bitumen in the mix design, the mechanics characteristics of materials are broadly cov-
ered through strains. For rutting, the compressive strain at the top of the soil subgrade
is considered whereas the tensile strain at the bottom of asphalt layer is used for the
estimation of fatigue performance.
Fundamentally, the strains estimated through layered-elastic theory is largely de-
pendent upon resilient modulus of various layers of pavement. One of the major as-
sumptions considered in the estimation of strain employs isotropic behavior of materi-
als. This assumption helps an easy estimation of materials’ properties using layered-
elastic analysis. However, since the granular layer is made up of various types of ma-
terials, anisotropic behavior remains prevalent instead of isotropic.
2

Research studies [1-3] have shown strong evidence of nonlinear cross-anisotropic

elastic behavior of granular materials. In literature, such behavior was modelled using
nonlinear cross-anisotropic modeling at different materials type, gradation, moisture
content, and stress levels. In another direction, several studies [4-5] were conducted to
incorporate the realistic material properties of the pavement layer and the moving traffic
load in the analysis of the flexible pavement using the finite element analysis. With
regard to the loading system, linear, nonlinear, static, and cyclic loading analysis were
also considered to simulate the realistic traffic loading. It was found that if pavement
designs are carried out assuming static loading and linear pavement materials, the de-
flection at the top of the subgrade becomes higher than the expected values when pave-
ment section with non-linear [7-9]. Several other research studies [10-12] utilized tri-
axial tests to assess the aggregate physical properties and their effects on the cross-
anisotropic behavior of unbound granular materials. These physical properties include
particle shape, form, angularity, texture, and gradation. It was highlighted that the ad-
vantage of the use of cross-anisotropy for the analysis of unbound granular bases is the
drastic reduction of bottom tensile stresses predicted by linear elastic analysis based on
the assumptions of isotropy [13-16].
Although a wealth of literature can be found that investigated the impact of aniso-
tropic behavior of granular materials, very limited studies can be found that explored
the impact of anisotropic behavior of granular materials on design and/or performance.
With this background, this study investigated the effect of anisotropic behavior of gran-
ular materials and quantify the effect on the overall design of the flexible pavement.
The scope of the study included (Figure 1):
• Conduct a comprehensive literature review pertinent to the material charac-
terization and design consideration of granular materials;
• Analyze the effect of linear and nonlinear stress-strain variation in Granular
Sub-base (GSB)layer;
• Simulate anisotropy behavior of the material through finite element model;
• Assess strains due to the anisotropic behavior using ABAQUS model; and
• Quantify the change in estimated design lives based on strains characteristics

Impact of Anisotropic Behaviour

Isotropic behaviour Anisotropic behaviour

Conventional Design
Non-linear Cross-anisotropy
IRC 37:2018 & IITPave
ABAQUS Modeling
Compute Strain
Compute Strain

Predict Performance
Predict Performance

Compare performance

Fig.1. Research methodology

 3

2 Theoretical Background and Relevant Literature

Nonlinear behavior is commonly characterized by stress-dependent resilient modulus.

In general, resilient modulus is assigned into finite element-based mechanistic pave-
ment analysis methods to predict the pavement responses such as stress, strain, and
deformation [17-21]. Further, analysis of a typical flexible pavement considering the
cross-anisotropy of granular base layer shows that the values of critical mechanistic
parameters (εv and εt) are higher than the isotropic material properties. With the varia-
tion of the degree of cross-anisotropy, the value of εv and εt increased, causing a reduc-
tion in rutting and fatigue lives [22-23].
Many research studies [22-24] found that the consideration of the anisotropy behav-
ior of granular materials shows a more accurate and more realistic estimation of the
critical mechanistic parameters instead of consideration on the isotropic behavior of
granular materials. In the design guidelines of the flexible pavement, IRC 37: 2018 [25]
the isotopic behavior of granular materials is considered for the analyses and design
purpose which make a less accurate design [24].

3 Finite Element Modeling

A finite element-based software ABAQUS® was used for modeling of the pavement
section. ABAQUS® model was prepared for the two different cases. In the first case,
the model was prepared for the simple isotropy case. In that case, the whole granular
layer shows the isotropic behavior. The critical strain values obtained from the analysis
of this model were used in the calibration of the new design parameters for the next
pavement model that considered the anisotropy in the granular layers. In the second
case, the pavement model is prepared for a new calibrated design parameter that shows
the anisotropic behavior of the granular layer. For that purpose, the whole granular layer
is divided into the numbers of layers of equal thickness. At first, the pavement section
was designed with the help of IRC 37:2018 [25] and critical strains were obtained. The
input parameters used in the design are summarized in Table 1.
Table 1. Critical Strain Values Obtained from IITPAVE and ABAQUS
ABAQUS Model: Sectional Area: (3000 x 3000) mm 2 IITPave
Thick
Modulus Poisson Vertical Tensile
CBR MSA Layer ness
(MPa) ratio Strain Strain
(mm)
Binder 3000 0.35 150
3.51x10 1.88x10
8% 30 Granular 199 0.35 410 -04 -04
Soil 67 0.35 -
ABAQUS Model: Sectional Area: (3000 x 3000) mm 2 ABAQUS
Binder 3000 0.35 150
1.94x10 9.12x10
8% 30 Granular 199 0.35 410 -04 -05
Soil 67 0.35 -
4

Next, a 3D pavement model was prepared in ABAQUS® as shown in Figure 2 (a).

The dimensions and other properties of pavement section were the same for the
ABAQUS model. Then, standard wheel load was applied and critical strain values were
estimated as illustrated in Figure 2 (b).

(a)

(b)
Fig.2. Finite Element Modeling of pavement: (a) model element, and (b) deformed
model characteristics

To improve the accuracy of the model, a strain calibration was conducted with re-
spected to the strains obtained from elastic layer analysis. With this, ABAQUS® model
was made possible to capture only the impact of non-linearity and cross-anisotropy
since all other design parameters of models remained same. Considering the calibrated
design parameters, a 3D pavement model was prepared in the ABAQUS®. In this
model, the thickness of the granular layer was divided into the number of layers of
equal thickness in vertical direction as shown in Figure 3 (a). Similar to the previous
model, standard wheel load was applied in the model and critical strains were deter-
mined as shown in Figure 3 (b).
 5

(a)

(b)
Fig.3. Finite Element Modeling accounting for non-linearity: (a) model elements in
which granular materials were divided, and (b) deformed model characteristics

4 Results and Analyses

4.1 Estimation of critical strain values

To compare the linear-elastic behavior and anisotropic behavior, it is deemed important

to idealize a pavement section. With this idealized section, isotropic and anisotropic
properties were assigned in ABAQUS® model and corresponding strains were esti-
mated. For this purpose, a pavement section was designed as per the IRC 37:2018 [25]
with 30 MSA design traffic volume and 8% CBR of subgrade soil. Next, critical strains
were estimated using the design equations provided in IRC 37:2018 [25] with 90%
reliability. After that, a set of trial thicknesses were calculated using modulus values
summarized in Table 1. Using IITPave software, tensile strain at the bottom of asphalt
layer (εt) and compressive strain at the top of subgrade (ε z) were estimated; these are
incurred strains of the pavement section with trial thickness. When the incurred strains
obtained from trial section were lower than the critical strains obtained from IRC
37:2018 [25] with 90% reliability, it was considered as design thickness. For all the
finite element modeling, these thicknesses of pavement section were idealized; and dif-
ferent material properties, such as isotropic and anisotropic section was assigned in
modeling to account for materials’ variability.
6

Next, a pavement section was modeled with the help of the ABAQUS®. In the pro-
cess, a simple model of two-layered pavement sections was created and standard wheel
load was applied on the model to simulate the critical strains of pavement. Here, critical
strains refer to tensile strain at the bottom of asphalt layer and compressive strain at the
top of subgrade layer. Then, to improve the accuracy of the ABAQUS model, it was
calibrated based on the critical strains obtained from IITPAVE software.

4.2 Critical strain and traffic level Estimation

The calibrated ABAQUS® pavement model was used to simulate non-linear elastic be-
havior. It is important to note that modulus of materials for non-homogeneity does not
consider a single value of modulus and thereby necessitating different variations in
moduli across the vertical direction. Even though no defined variation may be observed
for granular material, modulus was varied in defined format to illustrate the magnitude
of change and its associated impact on pavement performance. These variations were
carefully selected in order to capture gradual and non-consistent change in the proper-
ties of granular materials.
The resilient modulus of the granular sub-base layer was increased from 100 MPa at
the top and 200 MPa at the bottom in different variations. Since a modulus value of 200
MPa was selected for layered-elastic theory for granular layer (Table 1), similar mag-
nitude of modulus was considered here to make it comparable. With these initial and
final limits, the modulus was varied in four variations, namely: (i) linear variation (Fig-
ure 4(a)), (ii) quadratic variation (Figure 4(b)), (iii) exponential variation (Figure 4(c)),
and (iv) Logarithmic variation (Figure 4(d)).

(a) (b)

(c) (d)
Fig.4. Variation in Modulus: (a) Linear, (b) Quadratic, (c) Exponential, and
(d) Logarithmic
 7

Four separate models were prepared for each variation and critical strains were cal-
culated. Next, these critical strains obtained from each model were used to predict the
rutting and fatigue life as per IRC 37:2018 [25] design equation with 90% reliability. It
is important to recall that the dimensions and thicknesses of all these models were kept
same as summarized in Table 1. Critical strains obtained from all four variations and
associated rutting and fatigue lives are summarized in Table 2.

Table 2. Critical strains and estimated lives

Tensile Vertical Traffic Level (MSA)
Variation
Strain Strain Fatigue Rutting
Linear 1.97x10-04 2.96x-04 33.80 140.39
Quadratic 2.05x10-04 3.72x10-04 28.95 49.81
-04 -04
Exponential 1.99x10 3.63x10 32.49 55.66
Logarithmic 1.95x10-04 3.58x10-04 35.16 59.28

When the model findings are compared, deviation in performance and strain can be
quantified. The strain values summarized in Table 2 can be benchmarked with the strain
obtained from layered-elastic analysis. Furthermore, the rutting and fatigue perfor-
mance can be compared with 30 MSA design traffic since the pavement section was
idealized for 30 MSA design traffic. The major findings can be summarized as follows:
• Linear Variation: tensile strain for linear variation was more than the design
critical tensile strain. And vertical compressive strain was less than the design
critical vertical strain. The corresponding traffic level in MSA was more than
the design traffic (30 MSA). Thus, for the linear variation, the strength of the
granular materials was underestimated.
• Quadratic Variation: Strains for quadratic variation was found higher than
the design critical strain. In addition, the estimated rutting and fatigue life were
found to be lesser than 30 MSA. Hence, for the quadratic variation, the
strength of the granular materials was overestimated.
• Exponential Variation: Strains for exponential variation were more than the
design critical strain and the corresponding traffic level in MSA was more than
the design traffic. Thus, for the quadratic variation, the strength of the granular
materials was underestimated.
• Logarithmic: Strains for logarithmic variation were more than the design
critical strain and the corresponding traffic level in MSA was more than the
design traffic. Hence, for the quadratic variation, the strength of the granular
materials was underestimated.

As observed, nonlinearity in the granular layer had a significant effect on the rutting
and fatigue performance of flexible pavement. Comparison of the estimated traffic level
with the design traffic level showed the overestimation or underestimation of the
strength of the granular materials when anisotropy was considered. Expect the linear
variation in modulus, all other variations produced lower design traffic which was an
indication of overestimation of material properties. In practice, though the variation in
8

granular materials does not follow any particular pattern, the findings showed that any
variation in modulus other than linear distribution potentially reduce the performance
of pavement. This may lead to a serious concern since any variation of modulus actually
triggers the pavement to fail in prematurely. This is especially true when stress-depend-
ent non-linearity is exemplified by granular materials.

5 Research Significance

A significant contribution of this study was to analyze the effect of anisotropic behavior
of granular material in the design of flexible pavement. Different types of nonlinear
variation in resilient modulus at various depths of the granular layer was analyzed in
this study. This study was the first of its kind since it considered the anisotropic behav-
ior in the granular layer and its effect on the design traffic level. It is noteworthy that
the findings of the study set a strong platform towards advancing the state-of-the-art
and knowledge pertaining to the impact of non-linear behavior of materials on its pave-
ment performance.
• Anisotropic behavior of granular materials: Anisotropic properties are con-
sidered instead of isotopic granular materials properties using finite element
modeling. The model was calibrated, and then used in tandem with the lay-
ered-elastic theory to capture the anisotropic behavior of granular materials.
The methodology was simple but powerful enough to differentiate the changes
in material properties in terms of strains and performance.
• Non-linearity in the granular layer: The granular layer was divided into dif-
ferent layers of equal thickness. Resilient modulus values are increased from
bottom to top with different types of nonlinear variation. For each variation,
the strains at critical locations were simulated with the help of ABAQUS®
model, and then used to calculate the design traffic based on IRC 37:2018 [25].
• Estimation of traffic level: For each nonlinear variation, separate design traffic
was found and that traffic level was compared with the design traffic obtained
from layered-elastic theory. After comparisons, an overestimation of material
properties was observed if granular material does not remain isotropic. When
the variation in modulus was linear, the analysis of layer-elastic theory still
remains valid in terms of the performance. However, non-linear variation
caused an overestimation of material properties which can eventually lead to
premature failure of pavement.

6 Conclusions

The objective of this study was to investigate the impact of anisotropic behavior of the
granular material and quantify the effect on overall design of flexible pavements. An
idealized pavement section was used for understanding the impact of anisotropy
through finite element modeling. Pavement section designed with layered-elastic the-
ory was compared with the pavement section of anisotropic behavior. Based on the
 9

strains, the performance of pavement was estimated using mechanistic empirical de-
sign.
It was observed that any variation in modulus, expect linear increment overestimate
the material properties. If such overestimation is considered in pavement design, it may
potentially lead to premature failure of pavement. Certainly, this finding was extracted
based on the assumption that the performance equation of design guidelines remains
valid for the strain derived from anisotropic material properties. Even though the as-
sumption of performance prediction does not provide accurate measure of anisotropic
behavior of materials, it broadly addressed the impact of anisotropic materials proper-
ties. A similar methodology with finite element modeling can be implemented in con-
junction with the various other performance prediction models. More sophisticated ma-
terials properties can be used in future to account for stress-dependent non-linear prop-
erties of granular materials using advanced finite element modeling.

References

[1] R. S. Harichandran, M.-S. Yeh, and G. Y. Baladi, “MICH-PAVE: A nonlinear

finite element program for analysis of flexible pavements,” Transp. Res. Rec.,
no. 1286, 1990.
[2] R. G. Hicks and C. L. Monismith, “Factors influencing the resilient response
of granular materials,” Highw. Res. Rec., vol. 345, pp. 15–31, 1971.
[3] I. Al-Qadi, H. Wang, and E. Tutumluer, “Dynamic analysis of thin asphalt
pavements by using cross-anisotropic stress-dependent properties for granular
layer,” Transp. Res. Rec., no. 2154, pp. 156–163, 2010, doi: 10.3141/2154-16.
[4] Q. Liu and A. Shalaby, “Simulation of pavement response to tire pressure and
shape of contact area,” Can. J. Civ. Eng., vol. 40, no. 3, pp. 236–242, 2013.
[5] A. Nataatmadja, “On the response of granular unbound pavement materials
under repeated loading,” 1994.
[6] Y. H. Huang, “Pavement analysis and design,” 2004.
[7] L. Wang, L. R. Hoyos, J. Wang, G. Voyiadjis, and C. Abadie, “Anisotropic
properties of asphalt concrete: characterization and implications for pavement
design and analysis,” J. Mater. Civ. Eng., vol. 17, no. 5, pp. 535–543, 2005.
[8] IALA-AISM, “Guidelines for the design of footbridges,” IALA Guidel., no.
December, pp. 1–37, 2001, [Online]. Available:
http://trid.trb.org/view.aspx?id=873476.
[9] J. H. Oh, R. L. Lytton, and E. G. Fernando, “Modeling of pavement response
using nonlinear cross-anisotropy approach,” J. Transp. Eng., vol. 132, no. 6,
pp. 458–468, 2006, doi: 10.1061/(ASCE)0733-947X(2006)132:6(458).
[10] T. Pan, E. Tutumluer, and J. Anochie-Boateng, “Aggregate morphology
affecting resilient behavior of unbound granular materials,” Transp. Res. Rec.,
no. 1952, pp. 12–20, 2006, doi: 10.3141/1952-02.
[11] M. N. S. Hadi and B. C. Bodhinayake, “Non-linear finite element analysis of
flexible pavements,” Adv. Eng. Softw., vol. 34, no. 11–12, pp. 657–662, 2003,
doi: 10.1016/S0965-9978(03)00109-1.
10

[12] S. Underwood, A. H. Heidari, M. Guddati, and Y. R. Kim, “Experimental

investigation of anisotropy in asphalt concrete,” Transp. Res. Rec., vol. 1929,
no. 1, pp. 238–247, 2005.
[13] J. Uzan, “Characterization of Granular Material.,” Transp. Res. Rec., pp. 52–
59, 1985, doi: 10.1016/0148-9062(86)91013-2.
[14] A. Nataatmadja and A. K. Parkin, Characterization of granular materials for
pavements, vol. 26, no. 4. 1989.
[15] A. Adu-Osei, “Characterization of unbound granular layers in flexible
pavements. Report No.ICAR/502-3,” p. 260, 2000, [Online]. Available:
https://repositories.lib.utexas.edu/bitstream/handle/2152/35440/502-
3.pdf?sequence=2&isAllowed=y.
[16] S. H. Kim, D. N. Little, E. Masad, and R. L. Lytton, “Estimation of level of
anisotropy in unbound granular layers considering aggregate physical
properties,” Int. J. Pavement Eng., vol. 6, no. 4, pp. 217–227, 2005, doi:
10.1080/10298430500335244.
[17] M. Kim, E. Tutumluer, and J. Kwon, “Nonlinear Pavement Foundation
Modeling for Three-Dimensional Finite-Element Analysis of Flexible
Pavements,” Int. J. Geomech., vol. 9, no. 5, pp. 195–208, 2009, doi:
10.1061/(asce)1532-3641(2009)9:5(195).
[18] C. Johnson, “This is an Accepted Manuscript of an article published by Taylor
& Francis Group in,” vol. 53808, no. April, pp. 1–32, 2007.
[19] B. Ghadimi and H. Nikraz, “A comparison of implementation of linear and
nonlinear constitutive models in numerical analysis of layered flexible
pavement,” Road Mater. Pavement Des., vol. 18, no. 3, pp. 550–572, 2017.
[20] Y. Cai, A. Sangghaleh, and E. Pan, “Effect of anisotropic base/interlayer on the
mechanistic responses of layered pavements,” Comput. Geotech., vol. 65, pp.
250–257, 2015, doi: 10.1016/j.compgeo.2014.12.014.
[21] U. Seyhan, E. Tutumluer, and H. Yesilyurt, “Anisotropic aggregate base inputs
for mechanistic pavement analysis considering effects of moving wheel loads,”
J. Mater. Civ. Eng., vol. 17, no. 5, pp. 505–512, 2005.
[22] B. Beriha and U. C. Sahoo, “Analysis of flexible pavement with cross-
anisotropic material properties,” Int. J. Pavement Res. Technol., vol. 13, no. 4,
pp. 411–416, 2020, doi: 10.1007/s42947-020-0284-9.
[23] Y. Ozawa and J. Maina, “Influence of Cross-Anisotropy Material Behavior on
Back-Calculation Analysis of Multi-Layered Systems,” Analysis, no. July, pp.
27–34, 2008.
[24] W. J. V. der M. Steyn, J. W. Maina, and A. Repsold, “Comparison of isotropic
and cross-anisotropic analysis of pavement structures,” 2013.
[25] IRC 37:2018, "Guidelines for the design of flexible Pavement", Fourth
Revision, Indian Roads Congress, New Delhi, India.