Professional Documents
Culture Documents
by
Rong Luo
2007
The Dissertation Committee for Rong Luo Certifies that this is the approved version
of the following dissertation:
85B
Subgrade Shrinkage
87B
Committee:
Jorge A. Prozzi, Supervisor
Kenneth H. Stokoe, II
C. Michael Walton
Jorge G. Zornberg
Loukas F. Kallivokas
Krishnaswamy Ravi-Chandar
Subgrade Shrinkage
89B
by
97B
Dissertation
91B
Doctor of Philosophy
92B
August 2007
94B
To My Parents
Acknowledgements
98B
Subgrade Shrinkage
96B
Publication No._____________
The State of Texas has the most extensive network of surface-treated pavements
in the nation. This network has suffered from the detrimental effects of expansive soils in
the subgrade for decades. Longitudinal cracking on the Farm-to-Market (FM) network is
one of the most prevalent pavement distresses caused by volumetric changes of expansive
subgrades. Engineering practice has shown that geogrid reinforcement and lime treatment
can effectively reduce the reflection of longitudinal cracking on the pavement over
shrinking subgrade. However, little is known about the mechanism leading to the
propagation of the shrinkage cracks to the surface of the pavement. The use of geogrid
reinforcement and lime treatment is mostly based on empirical engineering experience
and has not been addressed in depth.
This dissertation research evaluates the stress field and constitutive models of the
subgrade soil subjected to matric suction change. The non-uniform matric suction change
in the subgrade is simulated by a thermal expansion model in a finite element program,
ABAQUS, to determine the shrinkage stresses in the subgrade soil and pavement
vi
structure. Numerical solution by the finite element analysis shows that the most likely
location of shrinkage crack initiation in the subgrade is close to the pavement shoulder
and close to the interface of the base and subgrade. Linear elastic fracture mechanics
theory is used to analyze the crack propagation in the pavement. Compared to the fracture
toughness of the pavement materials, the stress concentration at the initial shrinkage
crack tip is large enough to drive the crack to propagate further. When the shrinkage
crack propagates through the whole pavement structure, a longitudinal crack develops at
the pavement surface close to the pavement shoulder.
Based on the analysis of shrinkage crack propagation, this dissertation
investigates the mechanism of geogrid reinforcement and lime treatment. The geogrid can
significantly reduce the stress concentration at the crack tip if the geogrid is placed at the
bottom of the base. A geogrid with a higher stiffness further reduces the stress intensity
factor at the upper tip of the shrinkage crack. The lime treatment can improve the
mechanical properties of the expansive soil in several ways. The lime-treated soil has
lower plasticity index, higher tensile strength and higher fracture toughness. The possible
location of the shrinkage crack initiation is not in the lime-stabilized soil but in the
untreated natural soil close to the bottom of the lime-treated layer, where tensile stresses
exceed the tensile strength of the untreated soil. The shrinkage crack is less likely to
develop through lime-treated soil, which has increased fracture toughness. The
combination of geogrid reinforcement and lime treatment offers the most benefit for the
control of dry-land longitudinal cracking. In a pavement with a lime-treated layer, the
best place to install the geogrid is at the interface between the lime-stabilized layer and
the untreated natural soil. If using a geogrid with high stiffness, the Mode I stress
intensity factor may be reduced to a certain level that is lower than the fracture toughness
of the pavement material.
vii
Table of Contents
9B
ix
List of Tables
0B
List of Figures
1B
Figure 6.11 Comparison of Mode I Stress Intensity Factor in Single Model and Multiple
Crack Models .................................................................................................................... 96
Figure 7.1 Pavement Model with Geogrid Reinforcement and Lime Treatment (Model
7.1) .................................................................................................................................. 101
Figure 7.2 Pavement Model with Geogrid Reinforcement and Lime Treatment (Model
7.2) .................................................................................................................................. 102
Figure 7.3 Shrinkage Cracks in Pavement with Geogrid Reinforcement and Lime
Treatment ........................................................................................................................ 104
Figure 7.4 Mode I Stress Intensity Factor at Upper Crack Tip of Shrinkage Crack....... 114
Figure 7.5 Mode I Stress Intensity Factor at Upper Crack Tip of Crack No. i............... 115
Figure 7.6 Mode I Stress Intensity Factor at Lower Crack Tip of Shrinkage Crack ...... 116
xii
changes in the clay volume (Snethen et al., 1975). As a result, expansive soils exhibit
significant volume changes with the variation of the amount of present water.
3
3-15
10-40
70-80
65-180
50-840
Illite
Montmorillonite
/100g)
(milliequivalent
Capacity
10-20
meter/g)
(square
Exchange
Kaolinite
Clay Mineral
Specific Surface
Cation
(%)
35
81
161
1
27
251
29
61
344
limit
Liquid
(%)
index
Plasticity
(%)
limit
Liquid
Na+
104
38
(%)
index
Plasticity
K+
166
90
34
(%)
limit
Liquid
Cation
101
50
(%)
index
Plasticity
Ca2+
158
83
39
(%)
limit
Liquid
99
44
11
(%)
index
Plasticity
Mg2+
Expansive soils in foundations and subgrade can cause serious damage to houses,
buildings, roads and pipelines because of the soils expansion or shrinkage when
moisture levels change. Problems associated with expansive soils are common
worldwide, as have been reported in the United States, Australia, Canada, China, India,
Israel, and South Africa (Chen, 1988). Expansive soils are found in 20 percent of the
United States (Krohn and Slosson, 1980). Texas, Colorado and Wyoming have the most
severe degree of expansive soil occurrences.
In Texas, more than half of the total damage caused by expansive soils occurs on
highways and streets, which costs the Texas Department of Transportation (TxDOT)
millions of dollars to repair every year (Jayatilaka and Lytton, 1997). Longitudinal
cracking on the Farm-to-Market (FM) network is one of the most prevalent pavement
distresses due to the volumetric change of the expansive subgrade. This type of dry-land
crack initiates in the drying subgrade soil and reflects from the highly plastic subgrade
through the pavement structure (Sebesta, 2002). Pavement and geotechnical engineers
have for many years attempted to eliminate the dry-land cracking resulting from the
expansive subgrade.
A number of methods have been used to treat expansive soils, which can be
grouped into three categories: i) alteration of expansive material by mechanical, chemical
or physical means; ii) control of subgrade moisture conditions; and iii) geogrid
reinforcement. Lime stabilization is the most extensively used alteration for modifying
the expansive soils in the subgrade. The lime treatment thickness can vary from 0.25 m to
1 m. Other commercial stabilizers, for example, Roadbond EN1 and EMC Squared, have
also been used for treating the expansive soils (Rajendran and Lytton, 1997). These noncalcium stabilizers have been shown to increase the strength and stiffness of the treated
4
soil, reduce the swelling, decrease the permeability, and moderate the suction. Using
vertical barriers at the edge of the pavement is a typical method for controlling the
subgrade moisture conditions. Jayatilaka et al. (1997) found that installing impermeable
geomembranes as vertical moisture barriers in pavement sections could reduce the
moisture variation in expansive subgrade and then restrain pavement roughness. To date,
geogrid reinforcement combined with lime treatment is the most effective method to
prevent longitudinal cracking on Farm-to-Market (FM) roads caused by the shrinkage of
expansive subgrade. In particular, in the Bryan District (Texas) the geogrid is placed at
the interface of the cement-treated or lime-treated subbase and a 3 to 4 inch flexible base.
Based on the investigation of maintenance base repairs over expansive soils in the San
Antonio, Lufkin and Bryan Districts, Sebesta (2002) found that the geogrid-reinforced
pavement section in the Bryan District had the best observed performance of all the
stabilized sections with various treatments for longitudinal cracking.
However, despite the preliminary success of geogrid in limiting longitudinal
cracks, little is know about the mechanism leading to propagation of the longitudinal
cracks to the surface of the pavement. The use of lime and geogrid is mostly based on
empirical engineering experience to prevent the longitudinal cracks due to expansive
subgrade. Lime stabilization has been shown to reduce the plasticity index (PI) and swell
potential of the expansive soil. Lime-treated layers are usually found to be intact without
any shrinkage cracks, while multiple shrinkage cracks develop in the untreated natural
soil beneath the lime-treated layer. The mechanism of lime-stabilized soil preventing
shrinkage cracks has not been clarified, and the benefit of the lime treatment has not been
quantified. Significant research efforts have been spent on geogrid reinforced flexible
pavements subjected to traffic loading (Kuo et al., 2003; Kwon, et al., 2005; Tingle et al.,
2005). However, the use of geogrids to control environment-induced pavement distresses
5
has never been addressed in depth to date. A theoretical approach is desirable to analyze
the mechanism of the dry-land longitudinal crack development and to minimize this type
of crack by means of lime-treatment and geogrid-reinforcement based on current
successful experience.
1.3 RESEARCH OBJECTIVES
14B
for the development of an optimal pavement design method to minimize the dry-land
longitudinal cracks.
1.4 DISSERTATION OUTLINE
15B
pavement surface. The chapter also presents the sensitivity analysis conducted to study
the effect of geogrid properties on its reinforcement benefits.
Chapter 6 reviews the properties of lime-treated soil and illustrates that the
inclusion of lime-treated layer in the subgrade changes the location and propagation of
shrinkage cracks. This chapter discusses the development of single shrinkage crack and
multiple cracks in the untreated subgrade soil beneath the lime-stabilized layer.
Chapter 7 presents the combined benefit of geogrid reinforcement and lime
treatment. The ideal installation position of the geogrid is determined based on the
analysis of the shrinkage propagation. The stress concentration at the crack tips is
compared among the studied models with different geogrid properties and the number of
shrinkage cracks in the model.
Chapter 8 summarizes the main findings and addresses the contributions of the
proposed methodologies. It concludes by outlining a number of directions in which the
proposed methodologies could be further extended.
The equilibrium conditions for a saturated soil can be described by the effective
stress, ( u w ) , in which is the total stress, and u w is the pore-water pressure. This
stress variable has the following tensor form (Fredlund and Rahardjo, 1993):
12
13
11 u w
22 u w
23
21
31
32
33 u w
in which
u w = pore-water pressure;
9
12 = 21
(2.1)
23 = 32
(2.2)
31 = 13
(2.3)
Assuming that the saturated soil behaves as an isotropic and linearly elastic
material, the effective stress variable is used to formulate the constitutive relations with
respect to the generalized Hookes law. Equations (2.4), (2.5) and (2.6) give the elastic
constitutive relations in the x1 , x 2 , and x3 directions:
u
11 = 11 w ( 22 + 33 2u w )
E
E
22 u w
22 =
( 11 + 33 2u w )
E
E
33 u w
33 =
( 11 + 22 2u w )
E
E
(2.4)
(2.5)
(2.6)
where
10
Unlike saturated soils, an unsaturated soil has more than one independent stress
variable because of the presence of soil suction. Soil suction is a measure of a soils
affinity for water (Chen, 1988). In other words, suction is a parameter indicating the
intensity with which the soil will attract water. Generally, soil with lower water content
has higher soil suction. The soil suction, commonly called total suction, can be
quantified in terms of relative humidity. Total suction consists of two components: matric
suction and osmotic suction. Matric suction is derived from the negative water pressure
associated with capillary phenomenon. Osmotic suction arises from the soluble salts in
the soil water, which produce the osmotic repulsion forces.
The total suction and its components can be measured in laboratory and in the
field (Fredlund and Rahardjo, 1993). The matric suction can also be predicted
theoretically by solving the moisture diffusion equation that governs the matric suction
distribution in the soil body (Mitchell, 1979, 1980). Based on the theoretical solution of
the diffusion equation, a number of computer programs have been developed that are
capable of estimating the matric suction profile in the pavement subgrade during different
climate seasons (Gay, 1994; Jayatilaka, 1999; Lytton et al., 2004).
Due to the existence of soil suction, more than one independent stress variable is
used to describe the equilibrium condition and to formulate the constitutive equations of
an unsaturated soil. Fredlund and Morgenstern (1976) used two stress variables in the
stress analysis of the unsaturated soil: net normal stress, ( u a ) , and the matric suction,
(u a u w ) , in which
is the pore-water pressure. Consequently, the stress state of the unsaturated soil can be
expressed by two independent stress tensors:
11
11 u a
21
31
12
13
23
22 u a
32
33
u a
and
u a u w
0
0 .
u a u w
0
ua uw
The two stress tensors cannot be combined into one because they have different
constitutive relations which depend on the soil properties.
Assuming that the unsaturated soil is isotropic and linearly elastic, the constitutive
relations can be formulated in terms of the two stress state variables, ( u a ) and
(u a u w ) , as shown in Equations (2.7), (2.8) and (2.9) (Fredlund and Rahardjo, 1993):
22
33
11 u a
ua uw
E
E
H
ua
u uw
= 22
( 11 + 33 2u a ) + a
E
E
H
ua
u uw
= 33
( 11 + 22 2u a ) + a
E
E
H
11 =
( 22 + 33 2u a ) +
(2.7)
(2.8)
(2.9)
where
E = modulus of elasticity for the soil structure with respect to a change in the net
normal stress, ( u a ) ; and
H = modulus of elasticity for the soil structure with respect to a change in matric
suction, (u a u w ) .
Every constitutive equation for the unsaturated soil can be explained as an
extension of each corresponding constitutive relation for the saturated soil because of the
additional stress variable (matric suction) in addition to the normal stress.
Equations (2.7), (2.8) and (2.9) can also be written in incremental forms:
1
1
d 11 = d ( 11 u a ) d ( 22 + 33 2u a ) + d (u a u w )
E
E
H
12
(2.10)
1
1
d ( 22 u a ) d ( 11 + 33 2u a ) + d (u a u w )
E
E
H
1
1
= d ( 33 u a ) d ( 11 + 22 2u a ) + d (u a u w )
E
E
H
d 22 =
(2.11)
d 33
(2.12)
(2.13)
in which dV is the volume change of the unsaturated soil, V is the soil volume at
initial state, and d v is the incremental volumetric strain.
Let x1 be the transverse direction perpendicular to the vehicle travel direction on
the pavement, x 2 be the longitudinal direction which is the vehicle travel direction, and
x1
(Transverse Direction)
x2
(Vehicle Travel Direction)
x3
(Vertical Direction)
If the initial condition is considered after the subgrade construction when the
subgrade soil is intact without any cracks, the initial strains are zero in all three
directions. During the desiccation process of the soil in the pavement subgrade, the lateral
strains (the strain in horizontal directions, 11 and 22 ) remain zero before crack
13
initiation because of lateral constraint. The field data collected by Konrad et al. (1997)
confirmed that drying soils experience a restrained desiccation so that the lateral strains
were maintained zero until a crack initiated in the soil. As a result, the incremental
horizontal strains in both transverse ( x1 ) and longitudinal ( x 2 ) directions remain zero
before cracking, which means:
u
u uw
11 = 11 a ( 22 + 33 2u a ) + a
=0
E
E
H
ua
u uw
22 = 22
( 11 + 33 2u a ) + a
=0
E
E
H
(2.14)
(2.15)
This fact indicates that the straining is forced to be one-dimensional. If isotropic elasticity
remains applicable, 11 = 22 33 (Morris, et al., 1992). Therefore, Equation (2.14)
can be rewritten as:
u
u uw
11 = 11 a ( 11 + 33 2u a ) + a
=0
E
E
H
By rearranging Equation (2.16), the following equation can be reached:
( 33 u a ) E 1 (u a u w )
11 u a =
1
H 1
(2.16)
(2.17)
in which ( 11 u a ) is the incremental tensile stress in the transverse direction. Since the
pore-air pressure is atmospheric for most practical engineering problems, the net normal
stress in the vertical direction, 33 u a , equals to 33 by setting atmospheric pressure
zero. The total vertical stress, 33 , so-called the overburden pressure, is produced by the
self-weight of the soil and the pavement layers covering the soil. This stress can be
calculated by:
33 = a ha + b hb + s hs
(2.18)
in which
normal stress and matric suction can be determined by laboratory experiments. Morris et
al. (1992) presented a constitutive equation as:
dV
d v =
= C t d ( u a ) + C a d (u a u w )
V
(2.19)
in which
log10 f log10 f
i
i
V
= volumetric strain;
V
hi = initial value of matric suction;
(2.20)
CEC = (PL )
meq / 100 g
(2.21)
CEC = (LL )
meq / 100 g
(2.22)
1.17
0.912
where
(2.23)
(2.24)
The calculated Ac and CEAc are used to obtain a guide number of h in the Chart for
the Prediction of Suction Compression Index (Figure 2.2) developed by McKeen (1980).
17
The guide numbers in Figure 2.2 are h for soils with 100 percent fine clay. To acquire
the value of h for real soil, the guide number determined by Figure 2.2 is reduced by
multiplying the percent fine clay. Finally, the obtained suction compression index may be
corrected by Equations (2.25) and (2.26) to compensate for the different initial volume of
soil mass during a wetting or drying process (Lytton, 2004):
h ( swell ) = h e
(2.25)
h ( shrinkage ) = h e
(2.26)
Figure 2.2 Chart for the Prediction of Suction Compression Index (McKeen, 1980)
62B
For lime treated soils, Lytton (2004) proposed a method for estimating the
plasticity index (PI) and the liquid limit (LL) as shown in Equations (2.27) and (2.28):
9 % lim e
PI lim e treated = PI untreated
(2.27)
18
LLlim e treated =
PI untreated
+b
a
(2.28)
19
Group
0.83
11
II
0.81
14
III
0.73
20
IV
0.68
25
0.68
25
VI
0.68
25
20
Matric suction can be measured using filter paper in the laboratory as described in
American Society for Testing and Materials (ASTM) standard, D 5298-03 (2003), which
is a simple and economical method for the suction range from 10 to 100,000 kPa. In this
test method, the soil specimen is placed with filter papers in an airproof container for
seven days. This duration is sufficient to allow different vapor pressures inside the
container to reach equilibrium, including pore-water vapor pressure in the specimen,
pore-water vapor pressure in the filter paper, and partial water vapor pressure in the air.
Subsequently, a calibration relationship is developed between the filter paper water
content with soil suction based on the type of filter paper used and the test procedure.
Finally, the suction of the specimen can be determined using the measured mass of the
filter papers and the calibration relationship.
The axis-translation technique is another method to directly measure the matric
suction in the laboratory. This measurement was originally proposed by Hilf in 1956 for
both undisturbed and compacted soil specimens (Fredlund and Rahardjo, 1993). During
the test, a closed pressure chamber is used to contain the unsaturated soil specimen. A
pore-water pressure measuring probe connects a tube full of de-aired water and the soil
specimen. The water in the tube has a tendency to go into tension producing negative
water pressure, which is measured by a gauge. By increasing the air pressure in the
closed chamber, the water has a greater tendency to go into tension. Once equilibrium is
reached, the matric suction of the soil can be determined based on the difference between
the air pressure in the chamber and the measured negative water pressure.
The tensiometer is a device commonly used in the field to directly measure the
negative pore-water pressure in a soil. The tensiometer allows equilibrium to be achieved
between the soil and the measuring system. At equilibrium, the water in the tensiometer
21
has the same negative pressure as the pore-water in the soil. Currently, there are different
types of tensiometers available for use in the field (Fredlund and Rahardjo, 1993).
2.4.2 Theoretical Model of Matric Suction Prediction
43B
u (0, t ) = U e + U 0 cos(2nt )
(2.29)
in which
u (0, t ) = matric suction at ground surface, in pF (kPa=0.098110pF);
t = time in seconds.
To study the suction not only at the ground surface but along the depth of the soil,
Mitchell developed a model to estimate the suction u ( y, t ) at any time t and depth y :
n
n
u ( y, t ) = U e + U 0 exp
y cos 2nt
y
in which
y = soil depth;
w p
,
d c
w = water density;
d = soil dry density;
p = unsaturated permeability; and
c = inverse slope of log suction (in pF) vs. gravimetric water content.
22
(2.30)
The equilibrium suction in Equations (2.29) and (2.30) can be estimated for
different locations based on the Thornthwaite Moisture Index (TMI) (Wray et al., 2005).
As defined in Equation (2.31), TMI is a parameter introduced by Thorthwaite (1948) to
characterize the moisture balance in a specific location taking into account rainfall,
potential evapotranspiration and the depth of available moisture stored in the rotting zone
of the vegetation. The calculation procedure of TMI includes three steps: i) determining
monthly potential evapotranspiration; ii) allocating available water to storage, deficit and
runoff on a monthly basis; and iii) totaling monthly runoff moisture depth, deficit
moisture depth and evapotranspiration to obtain annual values.
100 R 60 DEF
TMI =
Ep
(2.31)
where
(2.32)
23
Figure 2.4 Thornthwaite Moisture Index Spatial Distribution in Texas (Wray, 1978)
64B
24
Figure 2.5 Variation of Soil Suction of Road Subgrade with Thornthwaite Moisture Index
(Wray, 2005)
65B
If the soil is under a flexible impermeable cover, e.g., flexible asphalt pavement,
the matric suction under the pavement center line is different from that under the
pavement edge (shoulder). Mitchell (1979) obtained the analytical solution of steady state
matric suction within the soil body under a flexible impermeable cover of length L . The
matric suction under the impermeable cover has an approximate relationship with the
matric suction at the cover edge:
u y ( x ) U e + (u y U e )
cosh
2a
L
cosh
4a
(2.33)
where
a = soil active zone depth, under which the soil matric suction has a constant
value of U e .
The horizontal matric suction profile can be predicted using Equation (2.33) based
on the vertical matric suction profile. Consequently, the matric suction distribution under
a flexible impermeable pavement is obtained at each steady matric suction state.
Based on Mitchells models, a number of computer programs have been
developed to predict the matric suction profiles in the pavement subgrade, such as
FLODEF (Gay, 1994), PRES (Jayatilaka, 1999), WinPRES (Lytton et al., 2004), and
SUCH (Wray et al., 2005). Lytton et al. (2004) presented matric suction data at
equilibrium, dry and wet conditions predicted by WinPRES in a number of highway
construction sites in Texas. They also showed the matric suction compression index
(SCI) for different layers of soils in the subgrade. Some of these data will be selected for
use in the proposed finite element models later in this dissertation.
2.5 SUMMARY
20B
This chapter has discussed the stress and strain state in saturated and unsaturated
soils as well as the volumetric change theory of unsaturated soils. Particular attention has
been paid to Lyttons model (Equation 2.20) because it provides a reasonable and
relatively simple relation between the volumetric strain and three measurable variables.
One variable, matric suction, can also be predicted by theoretical methods. This model
will be used in the following chapters to simulate the differential matric suction change in
the subgrade soil. The available data of matric suction and matric suction compression
index in the literature make is possible to simulate matric suction change. The matric
suction data will be used as the only load on the proposed pavement model. The loading
condition differentiates the proposed model from most traditional pavement models that
26
have traffic as the primary load. Chapter 3 will present the details of pavement model
construction, load simulation, and modeling results.
27
of the model, and Section 3.4 discusses the results and findings, followed by a summary
of the chapter in Section 3.5.
3.1 MODEL CONSTRUCTION
21B
x3 is the vertical direction (see Figure 2.1). The assumptions of plane strain are realistic
for long bodies: for example, a pavement that is infinitely long in the travel direction with
constant cross-sectional area subjected to loads that act only in the x1 or x3 directions
and do not vary in the x 2 direction.
The modeled pavement structure consists of an asphalt surface layer, a granular
base and a multi-layered subgrade. Each pavement layer is assumed to be homogenous,
isotropic, linearly elastic, weightless, and bonded to the underlying layer. Because of
symmetry, a half-wide (4 m) pavement is studied to reduce computation effort. The
thickness of the asphalt layer is 0.025 m, and the thickness of the base is 0.250 m. The
subgrade of a pavement section in Fort Worth (Texas) is selected for this analysis based
on the available data in the literature (Lytton et al., 2004). With a total depth of 4.5 m,
this subgrade consists of six layers with different soils; each layer has a specific matric
suction compression index (SCI). In order to apply proper boundary conditions, one more
layer of 1.5 m without suction change is added to the bottom of the subgrade to make the
subgrade a total depth of 6 m. The width of the subgrade soil is extended to 12 m for the
29
purpose of applying different model constraints that will be presented in later sections.
Figure 3.1 shows the details of the constructed pavement model in ABAQUS. In this
model, the soil under the pavement is defined as pavement subgrade, while the soil not
under the pavement (the soil under the edge CD in Figure 3.1) is defined as field soil.
Edge AF in Figure 3.1 is the centerline of the pavement. These definitions make the
following model description more clear.
The Youngs moduli of the asphalt, base and subgrade are assumed to be 2,500,
350 and 75 MPa, respectively. Since the moduli of different layers of the subgrade are
not available in the reference, a representative average modulus of 75 MPa is used for all
subgrade layers and for the field soil. The use of a constant modulus for subgrade and
field soil has a minimum effect on the accuracy level of the modeling results because the
elastic moduli of different layers in the subgrade have approximately the same order of
magnitude. The Poissons ratio is assumed to be 0.35 for every layer of the pavement
structure, the subgrade and the field soil.
30
31
(3.1)
Matric suction data and compression index to be applied to the pavement structure
are selected based on a previous study in Texas. The selected matric suction data were
predicted by the computer program WinPRES developed by Lytton et al. (2004). In order
to consider the most critical case (a long-term heavy rain followed by an extended dry
period), this study selects two steady state vertical matric suction profiles: one with
extremely low matric suction values, another with extremely high matric suction values.
Since the subgrade soil is under an impermeable asphalt surface layer, the matric suction
change in the soil under the pavement centerline is different from the soil under the
pavement shoulder. Generally, the closer the location is to the pavement centerline, the
less matric suction change in the subgrade soil is noted. Mitchells approach (Equation
(2.33)) is used to predict the horizontal matric suction profile under the pavement based
on the vertical matric suction profile. Therefore, the matric suction distribution over the
32
subgrade cross section is obtained under the flexible impermeable pavement at each
steady matric suction state. Table 3.1 shows the matric suction distribution in the soil in
extremely wet condition; Table 3.2 presents the matric suction distribution in the
subgrade in extremely dry condition.
Table 3.1 Matric Suction Distribution in Wet Subgrade Soil
48B
Depth
(m)
0.50
1.00
1.50
2.00
2.50
3.00
3.50
4.00
0.15
2.5587
2.5579
2.5566
2.5546
2.5521
2.5489
2.5449
2.5400
0.30
2.5800
2.5800
2.5800
2.5800
2.5800
2.5800
2.5800
2.5800
0.45
2.5800
2.5800
2.5800
2.5800
2.5800
2.5800
2.5800
2.5800
0.60
2.5800
2.5800
2.5800
2.5800
2.5800
2.5800
2.5800
2.5800
0.75
2.5800
2.5800
2.5800
2.5800
2.5800
2.5800
2.5800
2.5800
0.90
2.5640
2.5634
2.5624
2.5610
2.5591
2.5567
2.5537
2.5500
1.05
2.5480
2.5468
2.5448
2.5420
2.5382
2.5333
2.5273
2.5200
1.17
2.5507
2.5496
2.5478
2.5451
2.5416
2.5372
2.5317
2.5250
1.32
2.5533
2.5524
2.5507
2.5483
2.5451
2.5411
2.5361
2.5300
1.47
2.5533
2.5524
2.5507
2.5483
2.5451
2.5411
2.5361
2.5300
1.59
2.5533
2.5524
2.5507
2.5483
2.5451
2.5411
2.5361
2.5300
1.74
2.5560
2.5551
2.5536
2.5515
2.5486
2.5450
2.5405
2.5350
1.89
2.5587
2.5579
2.5566
2.5546
2.5521
2.5489
2.5449
2.5400
2.01
2.5587
2.5579
2.5566
2.5546
2.5521
2.5489
2.5449
2.5400
2.16
2.5587
2.5579
2.5566
2.5546
2.5521
2.5489
2.5449
2.5400
2.28
2.5587
2.5579
2.5566
2.5546
2.5521
2.5489
2.5449
2.5400
2.43
2.5613
2.5607
2.5595
2.5578
2.5556
2.5528
2.5493
2.5450
2.55
2.5640
2.5634
2.5624
2.5610
2.5591
2.5567
2.5537
2.5500
2.70
2.5640
2.5634
2.5624
2.5610
2.5591
2.5567
2.5537
2.5500
2.85
2.5640
2.5634
2.5624
2.5610
2.5591
2.5567
2.5537
2.5500
3.00
2.5640
2.5634
2.5624
2.5610
2.5591
2.5567
2.5537
2.5500
3.15
2.5640
2.5634
2.5624
2.5610
2.5591
2.5567
2.5537
2.5500
3.30
2.5640
2.5634
2.5624
2.5610
2.5591
2.5567
2.5537
2.5500
33
3.45
2.5667
2.5662
2.5653
2.5642
2.5626
2.5605
2.5580
2.5550
3.60
2.5693
2.5689
2.5683
2.5673
2.5661
2.5644
2.5624
2.5600
3.75
2.5693
2.5689
2.5683
2.5673
2.5661
2.5644
2.5624
2.5600
3.90
2.5693
2.5689
2.5683
2.5673
2.5661
2.5644
2.5624
2.5600
4.05
2.5693
2.5689
2.5683
2.5673
2.5661
2.5644
2.5624
2.5600
4.20
2.5693
2.5689
2.5683
2.5673
2.5661
2.5644
2.5624
2.5600
4.35
2.5693
2.5689
2.5683
2.5673
2.5661
2.5644
2.5624
2.5600
4.50
2.5747
2.5745
2.5741
2.5737
2.5730
2.5722
2.5712
2.5700
Depth
(m)
0.50
1.00
1.50
2.00
2.50
3.00
3.50
4.00
0.15
3.5795
3.6164
3.6790
3.7687
3.8878
4.0392
4.2268
4.4550
0.30
3.5315
3.5667
3.6262
3.7116
3.8250
3.9692
4.1477
4.3650
0.45
3.4836
3.5169
3.5735
3.6546
3.7623
3.8992
4.0687
4.2750
0.60
3.4383
3.4700
3.5237
3.6007
3.7030
3.8330
3.9940
4.1900
0.75
3.3983
3.4285
3.4797
3.5531
3.6507
3.7746
3.9282
4.1150
0.90
3.3583
3.3870
3.4357
3.5056
3.5983
3.7163
3.8623
4.0400
1.05
3.3183
3.3456
3.3918
3.4580
3.5460
3.6579
3.7964
3.9650
1.17
3.2810
3.3069
3.3508
3.4137
3.4972
3.6034
3.7349
3.8950
1.32
3.2463
3.2710
3.3127
3.3725
3.4519
3.5528
3.6778
3.8300
1.47
3.2144
3.2378
3.2775
3.3344
3.4100
3.5061
3.6251
3.7700
1.59
3.1824
3.2046
3.2423
3.2964
3.3682
3.4594
3.5724
3.7100
1.74
3.1531
3.1742
3.2101
3.2615
3.3298
3.4166
3.5241
3.6550
1.89
3.1237
3.1438
3.1778
3.2267
3.2914
3.3738
3.4758
3.6000
2.01
3.0944
3.1134
3.1456
3.1918
3.2531
3.3310
3.4275
3.5450
2.16
3.0704
3.0885
3.1192
3.1633
3.2217
3.2960
3.3880
3.5000
2.28
3.0491
3.0664
3.0958
3.1379
3.1938
3.2649
3.3529
3.4600
2.43
3.0304
3.0471
3.0753
3.1157
3.1694
3.2376
3.3221
3.4250
2.55
3.0118
3.0277
3.0548
3.0935
3.1450
3.2104
3.2914
3.3900
2.70
2.9931
3.0084
3.0342
3.0713
3.1206
3.1832
3.2607
3.3550
2.85
2.9718
2.9863
3.0108
3.0460
3.0927
3.1520
3.2255
3.3150
34
3.00
2.9505
2.9642
2.9874
3.0206
3.0648
3.1209
3.1904
3.2750
3.15
2.9318
2.9448
2.9668
2.9984
3.0403
3.0937
3.1597
3.2400
3.30
2.9105
2.9227
2.9434
2.9731
3.0124
3.0625
3.1245
3.2000
3.45
2.8918
2.9034
2.9229
2.9509
2.9880
3.0353
3.0938
3.1650
3.60
2.8759
2.8868
2.9053
2.9319
2.9671
3.0119
3.0674
3.1350
3.75
2.8599
2.8702
2.8877
2.9128
2.9462
2.9886
3.0411
3.1050
3.90
2.8439
2.8536
2.8701
2.8938
2.9253
2.9652
3.0147
3.0750
4.05
2.8279
2.8370
2.8525
2.8748
2.9043
2.9419
2.9884
3.0450
4.20
2.8146
2.8232
2.8379
2.8589
2.8869
2.9224
2.9664
3.0200
4.35
2.8012
2.8094
2.8232
2.8431
2.8695
2.9030
2.9445
2.9950
4.50
2.6866
2.6906
2.6972
2.7068
2.7195
2.7357
2.7557
2.7800
The matric suction change at every location of the subgrade from the wet
condition to the dry condition is then calculated based on the matric suction distributions
at the two steady matric suction states. The logarithm of the matric suction change in the
subgrade ( log10 (h f / hi ) ) is also computed and shown in Table 3.3. The subgrade soils
under the pavement centerline are assumed to have zero matric suction change. For the
field soil that is not under the pavement, the logarithm of the matric suction change is
considered uniform between two steady matric suction states because no impermeable
cover exists on the top of field soil.
Table 3.3 Logarithm of Matric Suction Change in Modeled Pavement Subgrade
50B
Depth
(m)
0.75
1.25
1.75
2.25
2.75
3.25
3.75
4.00
0.15
0.30
0.45
0.60
0.75
0.90
1.05
1.17
1.32
1.47
1.59
1.74
1.89
2.01
2.16
2.28
2.43
2.55
2.70
2.85
3.00
3.15
3.30
3.45
3.60
3.75
3.90
4.05
4.20
4.35
4.50
If the matric suction change is uniform in an unconstrained elastic soil body, the
resultant swelling or shrinkage occurs in such a way as to cause a cubic element of the
soil solid to remain cubic, while experiencing changes of length on each of its sides.
Normal strains develop in each direction without inducing any normal stress. In this case,
36
the matric suction change also does not produce shear strains or shear stresses. However,
if the soil body has a nonuniform suction change field, as studied in this research, or if
the soil expansion or shrinkage is prohibited from taking place freely because of
restrictions placed on the boundaries, even if the matric suction change is uniform, the
shrinkage stress will develop in the soil. Once the shrinkage tensile stress exceeds the
tensile strength of the soil, shrinkage cracks will initiate and develop in the subgrade soil.
As a result, the boundary conditions are directly related to the magnitude of the shrinkage
stresses as well as the development of the shrinkage cracks.
In the proposed pavement model shown in Figure 3.1, no constraint is assigned to
the pavement surface (edge AB ), the shoulder of the pavement (edge BC ), and the
surface of the field soil (edge CD ). Boundary conditions are specified at three
boundaries: pavement centerline (edge AF ), the bottom of pavement subgrade and field
soil (edge FE ), and the right vertical edge of the field soil (edge DE ). Because of
pavement symmetry, edge AF in the proposed model is not allowed to have horizontal
displacement. Since the model size in depth is large enough for the subgrade to assume
no significant deformation below 6 m, edge FE is specified zero displacement. At edge
DE , four types of model constraint are applied respectively in order to consider both
lateral confinement and possible shrinkage cracks in the field soil:
Third, the upper 2 m of edge DE is not constrained, and the rest of this
boundary is fixed in the horizontal direction, which is illustrated in Figure 3.4.
This case simulates a 2-meter-deep crack developing from the soil surface
downward at the right edge of the field soil.
38
39
Figure 3.2 Proposed Pavement Model with the First Model Constraint
40
Figure 3.3 Proposed Pavement Model with the Second Model Constraint
41
Figure 3.4 Proposed Pavement Model with the Third Model Constraint
42
Figure 3.5 Proposed Pavement Model with the Fourth Model Constraint
As stated in previous sections, the expansive soils not only have strains associated
with the displacement functions, but also have strains due to matric suction variations.
Assuming the subgrade soil in this research is linearly elastic and isotropic, the
volumetric change of the expansive soil is evenly distributed in transverse ( x1 ),
longitudinal ( x 2 ) and vertical directions ( x3 ). In other words, let the matric suction h of
an elastic isotropic body in an arbitrary zero configuration be increased by a small
amount, all infinitesimal line elements in the volume undergo equal shrinkage since the
body is isotropic. All line elements maintain their initial directions. According to
Equation (3.1), the strain components due to the matric suction change are:
hf
1
11 = 22 = 33 = h log10
3
hi
12 = 23 = 31 = 13 = 32 = 21 = 0
(3.2)
(3.3)
The matric suction induced strains can be superimposed to the stress induced
strains to give:
22
33
h
1
( 11 22 33 ) + 1 h log10 f
E
3
hi
1
h f
1
= ( 11 + 22 33 ) + h log10
E
3
hi
1
h f
1
= ( 11 22 + 33 ) + h log10
E
3
hi
11 =
(3.4)
(3.5)
(3.6)
Under the plane strain assumption, 22 = 0 . Therefore, Equations (3.4) and (3.6)
can be rearranged as follows:
(1 ) 11 33 1
hf
11 = (1 + )
h log10
E
3
hi
(1 ) 33 11
E
33 = (1 + )
h f
1
h log10
3
hi
43
(3.7)
(3.8)
h
E
[(1 11 ) + 33 ] E 1 h log10 f
11 =
(1 + )(1 2 )
1 2 3
hi
h
E
[(1 33 ) + 11 ] E 1 h log10 f
33 =
(1 + )(1 2 )
1 2 3
hi
(3.9)
(3.10)
There is an analogy between the matric suction variation in the soil and the
temperature change in a solid. Considering a change in temperature T ( x, y ) , the change
of length, L , of a small linear element of length L in an unconstrained body is
calculated by Equation (3.11):
L = LT
(3.11)
t = T
(3.12)
For the plain strain problem with 22 = 0 , the full stress-strain relations are as follows
(Ugural, et al., 1995):
(1 ) 11 33
+ T
11 = (1 + )
E
(3.13)
(1 ) 33 11
+ T
E
33 = (1 + )
(3.14)
(3.15)
(3.16)
In both the matric suction variation problem and the temperature change problem,
the body has strains associated with the displacement functions as well as strains due to
other causes (moisture differential or temperature variation). The constitutive equations
are similar for the two problems. When comparing the corresponding constitutive
44
h
equations of the two problems, the logarithm of matric suction change ( log10 f ) is
hi
equivalent to the temperature change ( T ), and negative one third of the matric suction
1
compression index h is equivalent to (the thermal expansion coefficient).
3
Because of the analogy between the matric suction variation in a soil and the
temperature change in a solid, in this dissertation research, the logarithm of matric
suction change is simulated by the equivalent temperature change using a thermal
expansion model in ABAQUS. The thermal expansion model in ABAQUS is able to
define thermal expansion effects in terms of thermal strains based on specified thermal
expansion coefficients (ABAQUS, 2006). The thermal expansion effects can be isotropic,
orthotropic or fully anisotropic. ABAQUS requires thermal expansion coefficients, ,
that define the total thermal expansion from a reference temperature, 0 , as shown in
Figure 3.6. The thermal strains are generated in ABAQUS according to Equation (3.17).
th = ( , f )( 0 ) ( I , f I )( I 0 )
in which:
= current temperature;
I = initial temperature;
f = current values of the predefined field variables; and
45
(3.17)
ABAQUS assumes that there is no initial thermal strain when the reference temperature
is not equal to the initial temperature. This assumption is enforced by the second term in
Equation 3.17 which represents the strain due to the difference between the initial
temperature, I , and the reference temperature, 0 .
In the pavement model proposed in this research, the initial temperature of the
subgrade is assumed to be zero, and the final equivalent temperature is the logarithm of
the matric suction change. Because the matric suction change varied from different
locations in the subgrade soil, the subgrade is partitioned into a number of grids. Each
grid is assigned a final equivalent temperature which is the logarithm of the matric
h
suction change ( log10 f ) at the corresponding location shown in Table 3.1. The
hi
1
thermal expansion coefficient in the simulation is h . The thermal expansion
3
coefficients used in this model have negative values because an increase in matric suction
results in shrinkage of the soil instead of expansion. In other words, the subgrade and
46
field soil in the proposed model behave as a Negative Thermal Expansion material,
which contracts upon heating rather than expanding.
3.3 FINITE ELEMENT MESH
23B
subgrade and the base. The difference in the four cases is the magnitude of 11 max . With
the first model constraint shown in Figure 3.2, the right edge of the field soil is not
allowed to move horizontally, and 11 max is around 0.30 MPa (Figure 3.7). When a 2meter-long top-down crack develops at the right edge of the field soil (the third model
constraint shown in Figure 3.4), the strain energy induced by the nonuniform matric
suction variation in the soil body is released by this crack to generate the crack surfaces.
In this case, 11 max decrease to approximate 0.28 MPa (Figure 3.9). If the length of this
crack increases to 6 m (the second model constraint illustrated in Figure 3.3), which
means the right edge of the field soil is allowed to move freely in the transverse direction,
more strain energy is released, and 11 max is found to be 0.22 MPa (Figure 3.8). When
the 2-meter-long crack develops closer to the pavement shoulder (4 m horizontally away
from the right edge of the field soil), which is the fourth model constraint shown in
Figure 3.5, 11 max is around 0.25 MPa (Figure 3.6).
These findings indicate that whether the field soil is intact or has shrinkage
cracks, the magnitude of 11 max is considerably larger than the soil tensile strength
reported in the literature (Ayad et al., 1997). According to the crack initiation criterion, a
crack will develop in the soil if the tensile stress exceeds the tensile strength of the soil.
Therefore, independent of the shrinkage cracks in the field soil, macro cracks tend to
initiate in the pavement subgrade close to the pavement shoulder and close to the
interface of base and subgrade. This matches the location of the observed longitudinal
cracks on in-service pavement sections.
After the onset of the shrinkage cracks in the pavement subgrade, fracture
mechanics theory will be used to determine whether or not the crack is stable in the next
chapter. In order to reduce calculation effort, only one type of constraint will be assigned
at the right edge of the field soil when studying the crack propagation. Of the four cases
48
studied with different model constraints, the most critical case with the largest 11 max is
the model with the first constraint in which the field soil is intact without any shrinkage
cracks. Consequently, in the analysis of crack propagation, the boundary conditions of
the proposed model will be as follows:
3.5 SUMMARY
25B
This chapter has identified the possible location of shrinkage crack initiation by
constructing a finite element model of pavement over shrinkage subgrade. The
distribution of the matric suction change is estimated based on the available matric
suction data and the empirical models in the literature. The non-uniform matric suction
change is simulated using a thermal expansion model in ABAQUS. Four different
boundary conditions are applied to the proposed model; these model constraints cover a
variety of possible conditions in engineering practice. The model constraints and the nonuniform matric suction change in the soil result in tensile shrinkage stresses in the
subgrade and the field soil. The largest tensile stress in the subgrade develops in the area
close to the pavement shoulder and close to the bottom of the base. This area is the most
likely for the initial shrinkage crack to develop. If the initial crack is able to propagate to
the pavement surface, a longitudinal crack will occur at the pavement surface close to the
shoulder. This finding agrees with the field observation in terms of the location of the
dry-land longitudinal cracks.
Based on the tensile stress distribution, the next chapter will investigate why and
how the crack is able to propagate by means of the linear fracture mechanics theory.
49
50
Figure 3.7 Transverse Stress Distribution in Pavement without Geogrid (First Model Constraint)
51
Figure 3.8 Transverse Stress Distribution in Pavement without Geogrid (Second Model Constraint)
52
Figure 3.9 Transverse Stress Distribution in Pavement without Geogrid (Third Model Constraint)
53
Figure 3.10 Transverse Stress Distribution in Pavement without Geogrid (Fourth Model Constraint)
The crack initiation criterion described in Chapter 3 determines the onset of the
shrinkage crack: a shrinkage crack will initiate in the soil if the tensile shrinkage stress
exceeds the tensile strength of the soil. After the crack initiation, the propagation of the
crack depends on a number of factors, including the loading condition, the specimen
geometry and the crack length. It is important to examine whether the initial shrinkage
crack is stable in the subgrade or if it will develop in one or both directions. The
progression of the initial shrinkage crack is critical to the development of the dry-land
longitudinal crack found on the pavement surface.
The aim of this chapter is to track the propagation of the initial shrinkage crack
and to investigate the mechanism of dry-land longitudinal crack generation. The chapter
begins with Section 4.1, which introduces the fundamentals of fracture mechanics. The
linear elastic fracture mechanics theory provides necessary theoretical support to the
analysis of crack propagation. Section 4.2 presents the fracture properties of the
pavement materials. Section 4.3 describes the finite element modeling of the crack
propagation process in the pavement structure. The chapter concludes with some
summary remarks in Section 4.4.
4.1 FUNDAMENTALS OF FRACTURE MECHANICS
26B
of the shrinkage cracks in this study, Mode I is assumed. For linear elastic crack tip field,
the stress and displacement solutions were developed by Irwin for each crack mode with
respect to rectangular coordinates and polar coordinates (Lawn, 1993; Paris and Sih,
1965). The Irwin crack-tip solution is not included in this discussion but is available in
the mentioned references. Since stress concentration occurs in the vicinity of the crack
tip, a stress intensity factor, K , is used to quantify the components of stress and
displacement at the crack tip. Each crack propagation mode has an associated stress
intensity factor, i.e. K I , K II , K III , respectively. The magnitude of the stress intensity
factors depends on the applied loading, crack length and specimen geometry.
55
The Irwin crack-tip solution is important for determining the crack propagation
from the energy release point of view. For a crack configuration in Figure 4.2, let U E be
the strain energy in the system and the crack length be a prior to extension. The strain
energy release may be presented in an integral form over the crack surfaces immediately
behind the crack tip per unit width of front:
a
1
( yy u y + xy u x + zy u z )dx , ( u = constant)
U E = 2
a +a 2
(4.1)
In Equation (4.1), the factor 2 indicates the displacement of the two crack surfaces; the
factor means the proportionality between tractions and corresponding displacements;
the stresses ij
correspond to
r = xa ,
a x a + a , = 0 ; and the
a u
(4.2)
Using the Irwin crack-tip solution and proceeding to the limit a 0 , the strains energy
release rate may be integrated as:
K 2 K 2 K 2 (1 + )
G = I + II + III
E
E
E
(4.3)
in plane strain.
2u
At energy equilibrium, the strain energy release rate, G , is equal to the surface
energy of the generated two crack surface, 2 s ( s is the surface energy of one crack
surface). When G = 2 s , the corresponding stress intensity factor, K c , is called the
fracture toughness of the material, which is a constant material property and can be
measured by experiment. As a result, the fracture criterion is: when the stress intensity
factor is larger than the fracture toughness of the specimen, the crack is unstable and will
propagate to release energy until equilibrium is reached; when the stress intensity factor
is smaller than the fracture toughness, the crack remains stable.
For a crack in a linear elastic body, the strain energy release rate ( G ) and the
stress intensity factor ( K ) can be quantified using a path-independent integral, which is
called J integral (Rice, 1968). The J integral is defined as:
57
J = Wdy T ds
x
(4.4)
where:
Ti = ij n j ;
u = the displacement vector;
In the linear elastic case, the J integral is identical to the strain energy release rate G
and can be used to determine the stress intensity factors. Therefore, the J integral is
widely accepted as a quasi-static fracture mechanics parameter for linear response.
4.2 TOUGHNESS OF PAVEMENT MATERIALS
27B
As indicated in Section 4.1., if the stress intensity factor is larger than the material
fracture toughness, the crack will propagate; otherwise the crack will be stable. The stress
intensity factor can be estimated by finite element modeling; however, fracture toughness
for soils and pavement materials (HMA, crushed stone, lime-treated soil, etc.) has not
been well defined to date.
Harison et al. (1994) used a ring test to measure the mode I fracture toughness
( K IC ) of two compacted soils from Kentucky. They found K IC varied significantly with
the soil type and water content. For the first test clay, K IC decreased from 0.200 to
0.020 MPa m as the water content increased from 3 percent to 23 percent. For the
second test clay, K IC varied from 0.028 to 0.005 MPa m as the water content
increased from 3 percent to 16 percent.
58
Konrad and Cummings (2001) researched the K IC of frozen base and subbase
soils in pavement with different ice content. The measured fracture toughness increased
with the increase of ice content and soil average grain size. For a given volumetric ice
content, the K IC of frozen crushed stone was found to be larger than that of the frozen
sand. At temperature -5C, the K IC of crushed stone samples with an average dry
density of 2070 kg/m3 and a porosity of 22 percent increased approximately linearly from
0.050 to 0.400 MPa m as the volumetric ice content increased from 6 percent to 14.2
percent; the K IC of the frozen sand with dry density varying from 1490 to 1690 kg/m3
increased from 0.040 to 0.700 MPa m when the volumetric ice content increased
from 8 percent to 28 percent.
Mobasher et al. (1997) tested the K IC of asphalt concrete and asphalt-rubber
mixture at different binder content and two temperatures (-1 C and -7 C). For most
specimens in their test, the measured K IC decreased with the increase of temperature,
but increased when the binder content increased. For asphalt concrete, the measured K IC
values changed from 0.770 to 1.170 MPa m ; for asphalt rubber, the K IC value was
from 0.600 to 1.060 MPa m .
Since the fracture toughness of soil and pavement materials reported in the
literature is not a constant but depends on a number of variables (temperature, material
property, etc.), it is difficult to precisely predict whether or not the initial crack will grow
by using the fracture criterion. However, in order to study the crack propagation process,
this study conservatively assumes the fracture toughness of each pavement material to be
constant and to be close to the lower value presented in the literature. The assumed
fracture toughness for asphalt, base and subgrade soil is 0.700, 0.050 and 0.040
MPa m , respectively. If the crack stress intensity factor is larger than the assumed
fracture toughness, the crack may not propagate but the probability that it will grow is
59
greater. This approach will also be used in Chapter 5 to demonstrate the benefit of
geogrid reinforcement.
4.3 CRACK PROPAGATION PROCESS
28B
Based on the finite element modeling of the pavement over shrinking subgrade,
the largest transverse tensile stress (which is higher than the soil tensile strength) in the
subgrade develops close to the pavement shoulder horizontally and at the interface of the
base and subgrade vertically. As a result, the area with the highest tensile stress is the
most likely area for development of the initial crack, so the initial crack is assumed to be
0.6 m horizontally away from the pavement shoulder. The vertical location of the initial
crack should not be at the top of subgrade because the high-modulus base tends to
prevent cracking and keep the subgrade soil intact. Therefore, a number of trial cracks are
placed at different vertical locations with a fixed crack length of 0.1 m. The stress
intensity factors are calculated and compared to determine the vertical location of the
initial crack that has the largest stress intensity factor.
For this part of the study, the FE program ABAQUS is also used to model the
crack and to calculate the J integral and stress intensity factors. The crack is modeled
as a seam with specified crack tips and crack fronts in the Interaction Module of
ABAQUS/CAE. The seam, defined as an edge in the 2D FE model constructed in the
prior sections, is originally closed but can open during the analysis. ABAQUS/CAE
places overlapping duplicate nodes along the seam when the mesh is generated. All the
elements in the FE model are eight node biquadratic plane strain (CPE8R) elements. By
using these collapsed second-order elements, a mesh singularity can be obtained at the
crack tips. For the purpose of creating a 1 / r singularity in strain at the crack tips, a
value of 0.25 is used for the midside node parameter, which moves the midside nodes on
the element sides adjoining the collapsed edge to the quarter points of the elements. In
60
addition, the element sides at the crack tips are collapsed with single-node-type
degenerate element control.
The J integral and stress intensity factors are evaluated in ABAQUS along
several contours. Each contour is a ring of elements completely surrounding the crack tip
from one crack face to the opposite crack face. ABAQUS is capable of calculating the J
integral of a number of different contours. The first contour integral is determined using
all elements within the crack front and one layer of element outside the crack front.
Additional contour integrals are computed by adding a single layer of elements to the
group of elements that are used to calculate the previous contour integral. ABAQUS
automatically finds the elements that form each ring contour at the crack tip. Each
contour provides an evaluation of the J integral and stress intensity factors. Although
the J integral is path independent, the J -integral estimates from different contours
may vary because of the approximate nature of the finite element solution. Therefore, a
finer mesh is necessary at the crack tips. However, even with a finer mesh at the crack
tip, the first few contour integrals may be inaccurate. To obtain a contour integral with a
high level of accuracy, more contours should be evaluated until the value of contour
integral stays approximately constant from one contour to the next.
Defining d as the vertical distance from the upper crack tip to the subgrade top,
the first trial crack has a d of 0.01 m; the second trial crack has a d of 0.02 m (0.01 m
lower than the first trial crack), and so on. The crack tip regions are partitioned to
facilitate the generation of focused meshes. In the Step Module of ABAQUS/CAE,
history outputs with 20 contours are requested for each crack tip to calculate the J
integral and the stress intensity factors. Mode I stress intensity factor ( K I ) is of the most
interest because it directly relates to the longitudinal cracking on the pavement. The stress
intensity factors of all trial cracks are calculated and compared, which are shown in Table
61
4.1. As the trial crack moves farther away from the subgrade top, the stress intensity
factors at both crack tips (upper tip and lower tip) increase first and then decrease. The 5th
trial crack (upper tip 0.05 m away from the subgrade top) has the largest stress intensity
factors at both crack tips, which indicates that this crack has the most strain energy
release rate to propagate. Therefore, the 5th trial crack is selected as the initial crack,
which is 0.6 m horizontally away from the pavement shoulder and the upper tip of which
is 0.05 m vertically away from the interface of the base and subgrade.
Table 4.1 Stress Intensity Factors of Trial Cracks
51B
d (m)
1
2
3
4
5
6
0.010
0.020
0.030
0.040
0.050
0.060
K I ( MPa m )
The stress intensity factor at the upper and lower tips of the initial crack are 0.110
and 0.108 MPa m , respectively, and both are larger than the assumed fracture
toughness of the subgrade soil (0.040 MPa m ). Consequently, the initial crack is
assumed to grow in both directions with an increment of 0.06 m. The 0.06 m increment is
selected to let the upper tip of the crack be 0.01 m advance into the base layer so that the
non-geogrid case can be compared with the geogrid case in later sections. This enlarged
crack (Stage 2 in Figure 4.3) has a total length of 0.22 m, and the stress intensity factors
at the upper and lower crack tips are 0.183 and 0.141 MPa m , respectively. The stress
intensity factor at the upper crack tip (0.183 MPa m ) is much larger than the
supposed fracture toughness of the base. Based on the fracture criterion, the crack
62
The progression of shrinkage cracks has been analyzed in this chapter for
investigating the mechanism of dry-land longitudinal cracks at the pavement surface.
The linear elastic fracture mechanics theory has been reviewed and used to model the
shrinkage crack in the subgrade. The stress concentration at the crack tip is evaluated in
terms of stress intensity factors. This parameter depends on the loading condition,
specimen geometry and crack length. Stress intensity factor at the crack tip can be
determined by the strain energy release rate. A higher strain energy release rate is usually
associated with a larger stress intensity factor.
The fracture properties of the pavement materials are critical for crack
development. If the fracture toughness of a material is larger than the stress intensity
factor at the crack tip inside the material, the crack will be stable; otherwise the crack will
propagate to release more strain energy. Because the fracture properties of pavement
materials have not been well documented, conservative values are assumed for the
fracture toughness of each pavement material used in this model based on the available
data in the literature. Compared to the stress concentration level at the shrinkage crack
63
tips, the fracture toughness of pavement materials is not large enough to keep the
shrinkage crack stable. As a result, the shrinkage crack develops through the pavement
structure to the pavement surface and then generates the longitudinal crack at the
pavement surface.
In order to control the dry-land longitudinal crack, the corresponding treatment
methods should be effective in one or both of the following: 1) increasing the fracture
toughness of the pavement materials; 2) decreasing the stress intensity factor at the
shrinkage crack tip. The benefit of geogrid reinforcement or lime treatment can be
attributed to their ability to make stress intensity factors smaller than the fracture
toughness of the pavement materials. The following chapter will examine in detail the
mechanism of how geogrid controls the dry-land longitudinal crack and will quantify
the benefit of geogrid reinforcement.
64
65
Figure 4.3 Stress Intensity Factors of Crack in Non-Geogrid Pavement (Unit: MPam0.5)
Geogrid is defined by the American Society for Testing and Materials (ASTM,
2006) as a geosynthetic formed by a regular network of integrally connected elements
with apertures greater than 6.35 mm (14 in.) to allow interlocking with surrounding soil,
rock, earth, and other surrounding materials to function primarily as reinforcement.
Geogrids are manufactured in a factory-controlled environment; they are packaged in
sheets, placed in a roll or carton, and finally transported to the sites. At the project sites,
the geogrid sheets are unrolled on the prepared surface, overlapped to each other to form
a continuous geogrid blanket, and often physically joined to each other. Field
observations show that the geogrids are effective for the control of dry-land
longitudinal cracks at the pavement surface (Sebesta, 2002). However, the function
mechanism of geogrid reinforcement is not clear, and the use of geogrid is based purely
on experience.
To fill the abovementioned gap in the application of geogrid in pavement, this
chapter investigates in Section 5.1 the reasons the geogrid prevents a shrinkage crack
from developing through the pavement. This section details how the geogrid reduces
stress concentration at the shrinkage crack tip using fracture mechanics principles.
Section 5.2 further models the pavement with a geogrid layer at the interface of the base
and subgrade. The stress intensity factors at the crack tips are calculated with the geogrid
reinforcement. A variety of geogrid stiffness values are used to examine the effect of
geogrid properties on the reinforcement benefit. Section 5.3 is a summary of this chapter.
5.1 MECHANISM OF GEOGRID REINFORCEMENT
30B
from developing and propagating due to the shrinkage of the subgrade soils. Because the
base material interlocks the subgrade soil through the geogrid apertures, the geogrid can
be assumed to be fully bonded with the pavement layer materials. Field measurements
and modeling studies in the literature showed the development of interface shear stresses
at the geogrid-aggregate interface and at the subgrade-geogrid interface under repeated
traffic loading (Perkins and Svano, 2004; Kwon et al., 2005). However, the assumption of
full bonded interface condition is appropriate in this study because it is a conservative
approach. As opposed to traditional reinforcement where the assumption of full bond is
unsafe, for crack propagation, slippage between the geogrid and the soil results in some
level of stress relaxation.
When the crack initiates in the subgrade and propagates upward, the crack may go
through the geogrid since the geogrid is not a physical barrier but has apertures.
However, the geogrid has to deform with the crack opening, and a small deformation of
the geogrid results in significant forces. Consequently, the geogrid applies these
holding forces directly to the crack faces, as illustrated in Figure 5.1. The forces tend to
restrain the crack from opening and therefore reduce the stress intensity factor at the
crack tip, which is the basic mechanism by which geogrid prevents crack development. If
there is interface sliding, that is, the Mode II crack develops at the subgrade-geogrid
interface, part of the strain energy will release along the interface. Therefore, the Mode I
stress intensity factor ( K I ) will decrease according to Equation (4.3), then the
desiccation crack is even less likely to develop toward the pavement surface. As a result,
the assumption of full bonded condition under-estimates the benefit of geogrid reducing
crack, which is conservative and on the safe side.
67
Asphalt
Vertical Direction
Base
Geogrid
P
Subgrade
68
In the finite element model constructed in ABAQUS, the geogrid is placed at the
interface between the base layer and subgrade as reinforcement. This modeling approach
agrees with current engineering practice in Texas of using geogrid to prevent cracks
propagating from the shrinking subgrade. The one-dimensional truss element is selected
to model the geogrid because the geogrid is a slender structure that supports loading only
along the horizontal direction but cannot resist bending. In the 2D finite element model,
the truss elements representing geogrid reinforcement are embedded in the host
pavement elements using the embedded element technique offered by ABAQUS. The
embedded truss elements have identical displacement to the host solid elements; that is,
the embedded elements are fully bonded to the surrounding materials. This technique
fairly simulates the condition whereby a crack goes through the geogrid without breaking
it.
The geogrid is given a fixed Poissons ratio of 0.5 and six different representative
stiffness values: 400, 800, 1,600, 3,200, 6,400 and 12,800 kN/m. The use of a range of
geogrid stiffness values is to cover a range of materials and to simulate the different
tensile strength of the reinforced versus un-reinforced granular layers. There are seven
cases in total in the ABAQUS modeling: no geogrid, geogrid with 400 kN/m stiffness,
geogrid with 800 kN/m stiffness, geogrid with 1,600 kN/m stiffness, geogrid with 3,200
kN/m stiffness, geogrid with 6,400 kN/m stiffness and geogrid with 12,800 kN/m
stiffness. The pavement structure and layer properties are the same in each case as
presented in Section 3.1. A crack is introduced to the pavement and subgrade; the size
and position of this crack are exactly the same as those of the crack at Stage 2 in Figure
4.3. The stress intensity factors at crack tips in each case are calculated and compared in
Figure 5.2. In this figure, Crack 1 is the crack developed in the pavement without
69
geogrid-reinforcement, and Crack 2 through Crack 7 are the cracks developed in geogridreinforced pavement with a geogrid stiffness of 400, 800, 1,600, 3,200, 6,400 and 12,800
kN/m, respectively. As can be seen from Figure 5.2, the crack in the pavement without
geogrid has a Mode I stress intensity factor ( K I ) of 0.183 MPa m at the upper crack
tip. The geogrid with a stiffness of 400 kN/m reduces K I by around 11 percent to 0.163
MPa m . As the geogrid stiffness increases from 400 to 12,800 kN/m, K I continues
to decrease significantly from 0.163 MPa m to 0.087 MPa m . Figure 5.3 shows
the trend of K I decreasing with the increase of geogrid stiffness. It is observed that the
geogrid reinforcement effect does not linearly increase with the increase of geogrid
stiffness. As the geogrid stiffness increases, the reduction rate of K I decreases first and
then increases.
5.3 SUMMARY
32B
This chapter has described the mechanism of the geogrid controlling dry-land
longitudinal crack at the pavement surface. Since the geogrid has apertures, the shrinkage
crack is able to propagate through the geogrid layer. Therefore, the geogrid has to deform
with the crack opening. Because the geogrid has large tensile strength and stiffness, a
small deformation of the geogrid produces considerable forces which are applied to the
two crack surfaces directly. These forces tend to close the crack and then reduce the
stress intensity factor at the shrinkage crack tip. When the reduced stress intensity factor
is less than the fracture toughness of the pavement materials, the shrinkage crack does not
propagate but remains stable in the subgrade. The geogrid with higher stiffness is able to
provide larger forces to the crack surfaces, and then produces more reduction in the stress
concentration at the crack tips. Therefore, a shrinkage crack is less likely to grow if a
geogrid with larger stiffness reinforces the pavement.
70
71
72
Figure 5.2 Stress Intensity Factors of Cracks in Geogrid-Reinforced Pavement (Unit: MPam0.5)
0.2
0.18
0.16
0.14
0.12
0.1
0.08
0.06
0.04
0.02
0
0
2000
4000
6000
8000
10000
12000
14000
Figure 5.3 Relationship between Mode I Stress Intensity Factor of Crack Tip in Base and
Geogrid Stiffness
70B
73
Lime treatment is the most extensively used chemical alteration method to modify
the swelling properties of expansive soils. Using lime to stabilize expansive subgrade
soils is one of the most effective methods to reduce longitudinal cracking of pavements
due to subgrade shrinkage. The mechanism of lime treatment controlling this type of
longitudinal crack is completely different from that of the geogrid reinforcement. The use
of geogrid aims at reducing the stress concentration at the shrinkage crack tips. On the
other hand, the objective of using lime to treat the subgrade soil is to increase the tensile
strength and fracture toughness of the soil. If the fracture toughness of the soil can be
increased to a level that is larger than the stress intensity factor at the crack tip, the
shrinkage crack cannot propagate through the lime-treated layer to reach to the pavement
surface. Another advantage is that the moisture sensitivity of the subgrade soils is
significantly reduced, which results is less volumetric changes.
This chapter reviews the properties of the lime-treated soil in Section 6.1. The
inclusion of lime in the soil significantly changes the mechanical properties of the natural
soil. The change of soil properties has considerable effect on the initiation and
propagation of the shrinkage crack in the subgrade soil. Section 6.2 describes a finite
element model of pavement with a lime-treated layer. The distribution of the shrinkage
tensile stress is examined in the pavement model in order to estimate the possible location
of shrinkage crack initiation. The development of shrinkage cracks is investigated in
Section 6.3. This section constructs a number of models with different numbers of
shrinkage cracks. In each model, the stress concentration at the crack tips is evaluated in
terms of stress intensity factor. A summary of this chapter is presented in Section 6.4.
74
75
(6.1)
This relationship is graphically illustrated in Figure 6.2, in which the percent swell was
defined as volume change of the soil when the moisture content increased from the
optimum moisture content to the saturation moisture content. With the reduction in
percent swell or volume change, the swell pressure was considerably decreased by adding
lime to the soil. Basma and Tuncer (1991) found that lime percentage and curing time
profoundly reduce the swell potential and swell pressure of expansive soils. Figure 6.3
demonstrates the influence of the amount of lime and the influence of curing time on the
reduction of swell pressure of lime-treated clay. In their one-dimensional swell tests, the
swell potential of a high PI clay with a swell pressure of 2,600 kPa was reduced to 1,700
76
kPa with 10 percent hydrated lime (immediately) and was further reduced to 0 kPa with
28 days of cure at only 4 percent hydrated lime.
Figure 6.2 Relationship between Plastic Index and Swelling (Seed et al., 1962)
72B
Associated with a reduction in the swell pressure and swell potential is the
decrease in matric suction compression index ( h ). When using Equation (2.27) to
estimate the reduced plasticity index of the lime-treated expansive soil, the reduced
matric suction compression index can be predicted by the empirical procedure developed
by McKeen (1980), which is stated in Chapter 2. For example, if a subgrade soil with an
original matric suction compression index ( h ) of 0.0313 is treated by 8 percent of lime,
h may decreases to 0.0156 (Lytton, 2004). When subjected to the same matric suction
change and model constraints, the soil with smaller h will have smaller maximum
tensile stress developing in the soil body than the soil with higher value of h .
77
Figure 6.3 Swell Pressure as a Function of Lime Content and Period of Curing for Irbid,
Jordan, Clay (Basma and Tuncer, 1991)
73B
increases the tensile strength of the expansive soils. The stabilized soil with higher tensile
strength is able to resist larger tensile stress developing in the soil. If the tensile stress
induced within the stabilized layer does not exceed the tensile strength, the shrinkage
crack will not initiate in the soil. Therefore, lime treatment provides the most desirable
benefit for the expansive subgrade: decreasing the tensile stress induced by suction
change as well as increasing the soils tensile strength at the same time.
Poissons ratio of lime-treated soil is a stress dependent property. Poissons ratio
falls in the range of 0.1 to 0.2 when the stress level is less than 50 percent of the ultimate
compressive strength. At higher stress levels, Poissons ratio is within the range of 0.2 to
0.3. A typical value of between 0.15 and 0.2 is usually used for Poissons ratio of limestabilized soil (TRB, 1987).
Engineering practice has shown that there is little development of shrinkage
cracks in well designed lime-treated expansive clays. Figure 6.4 shows the cross section
of an in-service pavement with a lime-stabilized layer. The material comprising the
bottom layer of this section is high PI clay; the middle section is a lime-treated layer; and
the top layer has granular materials. This picture clearly indicates that the shrinkage
cracks developed in the shrinking soil while the lime-treated layer still remained intact. In
this example, the lime-stabilized layer showed tensile strength high enough to resist
shrinking stress produced by matric suction change. The benefit of lime treatment is
quantified by finite element modeling of the pavement with lime-treated layer over
expansive soil, which is stated in the following section.
80
Figure 6.4 Shrinkage Cracks in High PI Clay Covered by Lime-Treated Layer (Courtesy
of Lytton and Scullion)
74B
modulus of 250 MPa and a tensile strength of 1 MPa. Poissons ratio used for the limestabilized layer is 0.2 in this model. Figure 6.2 illustrates the details of the pavement
model with lime-treated layers.
In Chapter 3, four different boundary conditions are applied to edge DE in the
pavement model without lime-treated layer, respectively. The first boundary condition
with zero horizontal displacement at edge DE is found to be the most critical, resulting
in the largest 11 max in the subgrade. This is reasonable because the more the material is
constrained, the larger the tensile stresses will be that develop. In the model with limetreated layer proposed in this chapter, only one boundary condition, which produces the
critical 11 max , is applied at edge DE in order to reduce computational effort. If lime
treatment can successfully prevent shrinkage cracks from developing in the subgrade in
the critical case, it will definitely be effective for the other cases with less strict boundary
conditions. The full boundary conditions for this model are:
used in the model in Chapter 3, which are shown in Table 3.1. Again, the logarithm of
matric suction change is simulated by an equivalent temperature change using the thermal
expansion model in ABAQUS. The thermal expansion coefficient used in the thermal
expansion model is negative one third of the matric suction compression index of each
layer. Since the matric suction change varies in the subgrade soils, the subgrade is
partitioned into a number of grids. The initial temperature of each grid is assigned to a
82
value of zero; the final temperature of every grid is the logarithm of matric suction
change at that location.
83
84
The purpose of using the same model constraints and loading in both models is to
determine how the lime treatment changes the distribution and magnitude of tensile
stresses induced by the same matric suction change. In the lime-treated layer, the
increased elastic modulus (from 75 MPa to 250 MPa) results in increased tensile stress.
On the other hand, the decrease of matric suction compression index contributes to the
decrease of tensile stress. The combined effect of increased elastic modulus and
decreased matric suction compression index is of special interest in this study.
To clearly illustrate the distribution of transverse tensile stresses in the upper part
of the subgrade soil and the field soil, the calculation results are displayed in Figure 6.6
and Figure 6.7, respectively. As can be seen from Figure 6.6, the largest transverse tensile
stress in the lime-treated layer, 11 max , develops at the interface of base and lime-treated
layer close to point C . The magnitude of 11 max is found to be approximately 0.38
MPa. Figure 6.7 shows another area with large transverse tensile stress in the field soil.
This area is within the lime-stabilized layer and close to point G at the field soil
surface. The largest tensile stress in this area is around 0.60 MPa. Since the tensile
strength of the lime-stabilized soil can be as high as over 1 MPa, the shrinkage crack has
a low probability of initiation in the lime-stabilized layer under the pavement when it is
subjected to the critical matric suction change.
However, tensile stresses are also found in the untreated subgrade soil and
untreated field soil. The largest transverse tensile stress in the untreated subgrade soil has
a magnitude of 0.22 MPa. It develops close to the vertical line at edge BC in Figure
6.5, and close to the interface of the lime-treated layer and the untreated soil. Transverse
tensile stresses also develop at the surface of field soil (edge GD in Figure 6.5) with a
range of 0.2 to 0.3 MPa. Shrinkage cracks have a high probability to initiate in these
areas with transverse tensile stresses. As a result, the inclusion of a lime-treated layer
85
changes the location of initial shrinkage cracks. Without lime treatment, shrinkage cracks
initiate directly under the base layer and propagate to the pavement surface if the
pavement is not reinforced by geogrid. When treating the upper layer of subgrade and
field soil with lime, the shrinkage cracks develop in the lower layer of the subgrade and
in the field soil that is not next to the pavement shoulder but further away from the
pavement shoulder. The lime-treated layer becomes a protection layer that keeps the
shrinkage cracks a certain distance away from the pavement.
The shrinkage cracks initiating in the subgrade are likely to propagate to the
pavement surface. Therefore, the fracture toughness of the lime-treated layer is an
important parameter whose magnitude determines whether or not the shrinkage crack
initiating in the untreated soil propagates through the pavement structure. Ideally, the
fracture toughness of the lime-stabilized soil is large enough to maintain the crack stable
in the untreated soil beneath the lime-treated layer, as shown in Figure 6.4. The possible
propagation of the shrinkage cracks is studied in the following section.
86
87
Figure 6.6 Transverse Stress Distribution in Pavement with Lime-Treated Layer (a)
88
Figure 6.7 Transverse Stress Distribution in Pavement with Lime-Treated Layer (b)
Since the untreated subgrade soil is subjected to large transverse tensile stresses
(up to 0.22 MPa), an initial shrinkage crack with a length of 0.1 m is introduced to the
untreated subgrade soil under the pavement. The horizontal distance between the initial
crack and the pavement shoulder (edge BC in Figure 6.5) is assumed to be 0.2 m. To
determine the critical vertical position of the initial crack, a number of trial cracks are
placed at different vertical locations, while the horizontal coordinate and length of the
trial cracks remain unchanged. The upper crack tip of the first trial crack is 0.01 m
vertically away from the interface of the lime-treated layer and the untreated layer. The
second trial crack is moved 0.01 m downward; the third trial crack is moved 0.02 m
lower than the first trial crack; and so forth until the critical vertical position is
determined.
A total number of 7 trial cracks are studied in terms of the stress intensity factors
at the crack tips. The results are summarized in Table 6.1, in which d is the vertical
distance between the upper tip of the trial crack and the interface of the lime-treated layer
and the untreated subgrade soil. The No. 6 trial crack is determined as the initial crack
because the stress intensity factor at the upper crack tip is found to be the largest.
Compared to the initial crack in the pavement model without lime-treated layer studied in
Chapter 4, the initial crack studied in this section shows smaller stress intensity factors at
both crack tips, but they are still larger than the assumed fracture toughness of the
subgrade soil (0.040 MPa m ).
89
Table 6.1 Stress Intensity Factors of Trial Cracks in Pavement with Lime-Treated Layer
52B
d (m)
K I ( MPa m )
0.010
0.076
0.083
0.020
0.081
0.085
0.030
0.083
0.086
0.040
0.084
0.086
0.050
0.086
0.086
0.060
0.089
0.083
0.070
0.086
0.082
K IC and the critical energy release rate J IC . They found that the cement content was the
primary controlling factor for toughness. With the standard compaction effort (ASTM
D1557), the stabilized soil with a cement content of 15 percent by weight had an average
90
K IC was 0.092 MPa m . Since both lime stabilization and cement stabilization are
typically used in Texas to improve the quality of soil in terms of strength and modulus,
the lime-treated soil in this study is assumed to have similar properties to those of the
cement-treated soil. Under this assumption, it is reasonable that K IC of the lime-treated
soil has the same order as K IC of the cement-treated soil.
Asphalt, 0.025 m, 2500 MPa
4m
B
C
0.200m
Lime-Treated Layer
0.010 m
0.750m
0.240 m
6m
Single Shrinkage Crack
untreated soil. In field observation, the initiation of macro cracks at the natural soil
surface has spacing between 0.020 m to 0.024 m at the soil surface (Konrad and Ayad,
1997). The differential shrinkage in the soil leads to the Mode II fracture or an in-plane
shear fracture mode where a crack propagates along the shear plane, as illustrated in
Figure 6.9.
Figure 6.9 Mode II Crack in Shrinking Soil (Konrad and Ayad, 1997)
76B
The increased number (more than one) of Mode I cracks and the existence of
Mode II cracks release more strain energy induced by the differential suction change.
When strain energy is released by multiple cracks simultaneously, the strain energy
release rate at every crack tip should be less than the release rate from a single crack. As
a result, the stress concentration should be relieved at each crack tip of multiple cracks.
To quantify the reduction in stress concentration, multiple Mode I cracks are introduced
to the model of the pavement with a lime-treated layer. The models with different
numbers of shrinkage cracks are studied individually. The models studied have 2 cracks,
3 cracks, 4 cracks, and 5 cracks, respectively. To differentiate the models, the model with
a single shrinkage crack is labeled Model 6.1; the model with two cracks is Model
6.2; the model with three cracks is Model 6.3; and so on. Figure 6.10 presents the
location and layout of the multiple shrinkage cracks in different models. In each model,
the shrinkage cracks are parallel to the initial crack and have the same length and vertical
92
location. The distance between two neighboring cracks is 0.2 m. The Model I stress
intensity factors at crack tips are calculated in every model. The results are summarized
in Table 6.2.
Asphalt, 0.025 m, 2500 MPa
4m
B
G
Lime-Treated Layer
0.200m
0.200m
0.750m
ii i
6m
Multiple Shrinkage Cracks
4m
B
0.200m
Lime-Treated Layer
0.200m
0.200m
iii
0.750m
ii i
6m
Multiple Shrinkage Cracks
4m
B
0.200m
Lime-Treated Layer
0.200m
0.200m
iii
ii i
0.750m
iv
6m
Multiple Shrinkage Cracks
4m
B
0.200m
Lime-Treated Layer
0.200m
0.200m
iii
0.750m
ii i iv v
6m
Multiple Shrinkage Cracks
Label of
Studied Model
Number of
Cracks
Crack No.
6.1
6.2
6.3
6.4
6.5
K I ( MPa m )
Upper Crack
Lower Crack
Tip
Tip
0.177
0.116
0.158
0.108
ii
0.132
0.075
0.154
0.097
ii
0.111
0.061
iii
0.146
0.095
0.125
0.082
ii
0.108
0.057
iii
0.144
0.095
iv
0.149
0.092
0.122
0.079
ii
0.107
0.057
iii
0.143
0.094
iv
0.121
0.072
0.150
0.091
94
The Mode I stress intensity factors of the cracks are compared among all studied
cases with different numbers of shrinkage cracks, as shown in Figure 6.11. The single
crack has a larger K I at both crack tips than any one of the other cracks. When a second
crack develops in the subgrade, K I of Crack No. i decreases at both crack tips. As more
cracks develop in the subgrade, K I of Crack No. i continues to decrease because part of
the strain energy is released by the other cracks. The same trend applies to the other
cracks. For example, K I of Crack No. ii decreases when the third crack develops in the
subgrade. For each crack, the inclusion of an additional crack results in different levels of
reduction in K I , which means the decrease rate of K I is not constant. In addition, the
inclusion of a new crack has a different impact on every crack. When introducing Crack
No. iv, K I of Crack No. i decreases significantly at both crack tips, while Crack No. iii
and Crack No. ii have only a slight reduction in K I . The reason is that the decrease of
K I is dependent on the location of the new crack. The initiation of a new crack will
considerably reduce K I of its neighboring crack but will have little effect on the cracks
a certain distance away from the new crack. In other words, if the new crack develops
next to an existing crack, the existing crack will experience a significant reduction in
K I . Figure 6.11 gives good examples: the appearance of Crack No. v decreases the K I
of Crack No. iv significantly but has little impact on Crack No. iii; Crack No. ii has a
clear reduction in K I because of the occurrence of Crack No. iii, while Crack No. ii has
very slight change in the magnitude of K I with the initiation of Crack No. iv.
The development of more shrinkage cracks releases more strain energy and
reduces the stress concentration at the tips of existing cracks. However, the reduced K I
is still in the same order of the fracture toughness of the lime-stabilized layer. This fact
indicates that it is still possible for shrinkage cracks to propagate through the lime-treated
layer. To further reduce the possibility of crack propagation, a geogrid will be placed at
95
the interface of the lime-treated layer and the untreated soil. The geogrid may further
decrease the stress concentration at the crack tips. Therefore, the chemical method (lime
treatment) and the mechanical method (geogrid reinforcement) will be combined to
prevent the shrinkage cracks from developing upward to the pavement surface. The
modeling and analysis results of the model with both treatments will be presented in the
following chapter.
0.20
Upper Crack Tip
Lower Crack Tip
0.18
0.16
0.14
0.12
0.10
0.08
0.06
0.04
0.02
6.2
6.1
6.4
6.3
6.5
0.00
i
ii
ii
iii
ii
iii
iv
ii
iii
iv
Crack No.
Figure 6.11 Comparison of Mode I Stress Intensity Factor in Single Model and Multiple
Crack Models
78B
96
6.4 SUMMARY
36B
This chapter has investigated the mechanism of lime treatment controlling dryland longitudinal crack at the pavement surface. The lime treatment can significantly
change the properties of the natural soils through the processes of modification and
stabilization. The lime-treated soil has lower plasticity index, smaller matric suction
compression index and higher tensile strength. The tensile shrinkage stress developing in
the lime-treated layer cannot exceed the tensile strength of the lime-treated soil with
proper design of lime treatment. As a result, shrinkage cracks have lower probability of
developing in the lime-stabilized layer.
However, shrinkage cracks are likely to initiate in the untreated soil beneath the
lime-treated layer because of the development of large tensile stress in the untreated soil.
The lime-treated layer is likely to make the shrinkage cracks stable because the fracture
toughness has been considerably increased by the inclusion of lime. The theoretical
calculation shows that the stress intensity factor at the shrinkage crack tips is of the same
order of the fracture toughness of the lime-treated layer. If multiple shrinkage cracks are
present in the subgrade, each crack tip has less stress concentration compared to the
single shrinkage crack case. In other words, the occurrence of multiple shrinkage cracks
reduces the stress intensity factor at every crack tip. This fact further reduces the
probability of shrinkage crack propagating toward the pavement surface.
Since the geogrid can reduce the stress concentration at the crack tip, and the lime
treatment is able to increase the fracture toughness of the soil, it is desirable to combine
the two methods to be more effective for the control of the dry-land longitudinal crack.
The design with both geogrid reinforcement and lime treatment provides a safer and more
conservative approach for the pavement structure over expansive subgrade. The next
97
chapter will discuss the proper design and the benefit of combining geogrid
reinforcement and lime treatment.
98
level at which the cracks will remain stable. However, if the upper part of the subgrade
soil is treated with lime, the lime-stabilized layer changes the tensile stress distribution in
the subgrade and also changes the potential location of crack initiation. As a result, the
geogrid may not be effective for the control of shrinkage cracks if it is still placed at the
interface of base and subgrade. Three models are constructed in ABAQUS to study the
effects of the geogrid on the stress concentration at the crack tips when the geogrid is
placed at the interface of the base and the lime-treated layer. In all three models, the
stiffness of the geogrid is assigned a value of 400 kN/m.
The first model is labeled Model 7.1, which has a shrinkage crack with the
same location and length as the one in Model 6.1 (Figure 6.8) in Chapter 6. A geogrid is
placed at the interface of the base and the lime-treated soil (edge HC in Figure 7.1
which shows the details of this model). The modeling technique used for the geogrid in
Chapter 5 is applied in this model: the geogrid is modeled using one-dimensional truss
elements, which are embedded in the host pavement elements. The stress intensity factors
are calculated at the crack tips (see Table 7.1). When comparing K I of the crack in
Model 6.1 and that in Model 7.1, the magnitude of stress intensity factors does not
change according to the inclusion of the geogrid due to the fact that the geogrid provides
most reinforcement when the crack propagates through the geogrid and results in the
deformation of the geogrid. This fact indicates that in the pavement with a lime-treated
layer, placing a geogrid at the interface of the base and the lime-stabilized soil has little
effect on the stress intensity factors at both crack tips.
100
4m
A
Geogrid
0.200m
Lime-Treated Layer
0.010 m
0.750m
0.240 m
6m
Single Shrinkage Crack
Figure 7.1 Pavement Model with Geogrid Reinforcement and Lime Treatment (Model
7.1)
79B
The second model is labeled Model 7.2, which has a crack with a final length of
1.74 m. This model is constructed based on the assumption that the shrinkage crack
(Crack No. i in Model 7.1) is not stable in the subgrade and propagates through the limetreated layer. The purpose of constructing this model is to determine if the geogrid is able
to control the shrinkage crack propagation when the lime-stabilized layer cannot prevent
the crack from developing through the layer. Crack No. i in Figure 7.1 is assumed to
grow 0.75 m in both directions. Since the initial crack has a length of 0.24 m, the crack in
Model 7.2 has a total final length of 1.74 m. The upper crack tip is within the base layer
and has a vertical distance of 0.01 m from the interface of the base and the lime-treated
layer. The detailed layout of this model is shown in Figure 7.2. The calculated K I is
0.833 MPa m at the upper crack tip and is 0.287 MPa m at the lower crack tip
(see Table 7.1). These values are so large that the crack is unlikely to remain stable based
on the typical fracture toughness of the base. Compared to Stage 2 in Figure 4.3, this
crack is likely to develop upward to the pavement surface. This result suggests that the
geogrid reinforcement is not able to control the shrinkage crack if the geogrid is placed at
the interface of the base and the lime-treated layer.
101
4m
B
H
Geogrid
0.01 m
Lime-Treated Layer
0.750m
i
6m
1.74 m
Figure 7.2 Pavement Model with Geogrid Reinforcement and Lime Treatment (Model
7.2)
80B
Even though the geogrid reinforcement cannot stop the crack in Model 7.2, the
geogrid may still reduce the stress concentration at the crack tips to some extent. To
quantify the reduction in K I provided by the geogrid, a third model (Model 7.3) is
constructed which has a crack with the same length and location as the one in Model 7.2,
but the model does not include a geogrid. The computed K I values are shown in Table
7.1. Comparing the K I values in Model 7.2 and Model 7.3, the geogrid reduces the
stress concentration at the upper crack tip, but the reduction in K I is not significant
enough to keep the crack stable.
Table 7.1 Mode I Stress Intensity Factor of Shrinkage Cracks in Pavement with Geogrid
Reinforcement and Lime Treatment
54B
K I ( MPa m )
Upper Crack
Lower Crack
Tip
Tip
0.177
0.116
Label of
Studied Model
Number of
Cracks
Crack No.
7.1
7.2
0.833
0.287
7.3
i
102
0.928
0.288
Based on the calculation results of the three models (Models 7.1 through 7.3), the
geogrid is not able to control shrinkage cracks if the geogrid is installed at the interface of
the base and the lime-treated layer. Instead, the geogrid should be placed close to the
location where the shrinkage crack initiates. When the crack is at the initiation stage, the
stress concentration is lower, so the geogrid has a better chance to reduce K I to a level
that is less than the fracture toughness of the material. Consequently, the ideal location to
install the geogrid should be at the interface of the lime-treated layer and the untreated
soil.
7.3 MODELING OF PAVEMENT WITH GEOGRID AND LIME TREATMENT
38B
With the determination of the best location of geogrid in the pavement with a
lime-treated layer, pavement models are constructed in ABAQUS with a lime-treated
layer and a geogrid layer. The material properties of the asphalt, base and lime-treated
soil remain the same as the previous models in Chapter 6. The geogrid is placed at the
interface of the lime-treated layer and the untreated soil. Five models with different
numbers of shrinkage cracks are studied, labeled as Model 7.4, Model 7.5, Model 7.6,
Model 7.7 and Model 7.8, respectively. Figure 7.3 shows the detailed configuration of the
pavement structures of the five models. Model 7.4 is similar to Model 6.1, which has a
single shrinkage crack. The only difference between Model 7.4 and Model 6.1 is that a
geogrid layer is included in Model 7.4 at the bottom of the lime-treated layer (edge IJ
in Figure 7.3). The width of the geogrid is 4 m in the transverse direction. The same
difference exists between Model 7.5 and Model 6.2, between Model 7.6 and Model 6.3,
between Model 7.7 and Model 6.4, and between Model 7.8 and Model 6.5. Therefore, the
benefit of including a geogrid layer in a lime-treated pavement can be found by
comparing K I between two corresponding models.
103
4m
A
0.200m
Lime-Treated Layer
0.010 m
Geogrid
I
0.750m
0.240 m
i
J
6m
Single Shrinkage Crack
4m
A
Lime-Treated Layer
0.200m
Geogrid
0.750m
0.200m
ii i
J
6m
Multiple Shrinkage Cracks
4m
A
0.200m
Lime-Treated Layer
0.200m
Geogrid
0.200m
iii
0.750m
ii i
J
6m
Multiple Shrinkage Cracks
4m
A
0.200m
Lime-Treated Layer
0.200m
Geogrid
0.200m
iii
ii i
0.750m
iv
J
6m
Multiple Shrinkage Cracks
4m
A
0.200m
Lime-Treated Layer
0.200m
0.200m
iii
0.750m
ii i iv v
J
6m
Multiple Shrinkage Cracks
In each of the five models (Model 7.4 to Model 7.8), the geogrid is assigned
stiffness values of 400, 800, 1,600, 3,200, 6,400 and 12,800 kN/m in order to study the
effect of the geogrid stiffness on the lime-treated pavement. Therefore, every model has
six sets of K I values, and each set of K I values corresponds to specific geogrid
stiffness. For example, Model 7.4 with a single shrinkage crack is analyzed six times in
ABAQUS; in each analysis, the geogrid in Model 7.4 has a different stiffness value, from
400 to 12,800 kN/m. All the calculated stress intensity factors are categorized into six
105
tables, Tables 7.2 to 7.7, according to the stiffness value of the geogrid. Table 7.2 shows
all K I values of every model with a geogrid stiffness of 400 kN/m; Table 7.3 illustrates
K I values of all models with a geogrid stiffness of 800 kN/m; and so on.
The inclusion of geogrid significantly reduces the Mode I stress intensity factors
at the upper crack tips of Crack No. i, ii and iii in Model 7.4 to Model 7.8. A geogrid with
higher stiffness provides more reduction in K I values at the upper tips of Crack No. i, ii
and iii. Figure 7.4 summarizes K I values at the upper tips of cracks in Model 7.4 to
Model 7.8. In this figure, different markers represent different stiffness levels of the
geogrid. The geogrid stiffness has a considerable effect on K I at the upper crack tip of
Crack No. i, ii and iii in all five models. However, the effect of the geogrid stiffness on
K I at the upper tip of Crack No. iv is not constant in Models 7.7 and 7.8; K I at the
upper tip of Crack No. v changes slightly with the increase of the geogrid stiffness. The
possible reason is that Crack No. iv is placed at the right end of the geogrid (at point J
in Figure 7.3), and Crack No. v is not reinforced by the geogrid. Therefore, the geogrid
has a marginal effect on these two cracks.
When the geogrid has an extremely large stiffness (12,800 kN/m), increasing the
number of shrinkage cracks does not have a significant impact on K I at the upper crack
tips. Figure 7.5 shows K I at the upper tip of Crack No. i in Model 7.4 to Model 7.8. At
low geogrid stiffness level, the development of Crack No. ii and Crack No. iv, which are
next to Crack No. i, results in significant reduction in K I . In contrast, at high geogrid
stiffness level (12,800 kN/m), the development of additional shrinkage cracks cannot
significantly reduce K I at the upper crack tip of Crack No. i. Figure 7.5 also indicates
the nonlinear effect of geogrid stiffness on reducing K I at the upper crack tip of Crack
No. i. At lower geogrid stiffness level, the additional stiffness significantly reduces K I ,
106
while at high geogrid stiffness level, the additional stiffness provides only marginal
benefit on reducing K I .
The geogrid has little effect on K I at the lower crack tip of all studied shrinkage
cracks, as illustrated in Figure 7.6. Therefore, the shrinkage cracks are likely to propagate
downward into the deeper soil. According to empirical evidence, the shrinkage crack
under the lime-treated layer may develop downward as deep as 2 m.
7.3 SUMMARY
39B
This chapter has discussed the combined benefit of geogrid reinforcement and
lime treatment for preventing shrinkage cracks from developing through the pavement
structure. Based on the theoretical calculation, if the upper part of the expansive subgrade
is treated by a certain percentage of lime (or a combination of lime and cement), the best
place to install the geogrid should be at the interface of the lime-treated layer and the
untreated subgrade soil. At this location, the geogrid can provide the most benefit to
control the propagation of the shrinkage crack. The overall reduction in stress
concentration at the shrinkage crack tips is significant when both geogrid and limetreated layer are included into the pavement model. The stiffness of the geogrid has an
important effect on the stress intensity factors at the crack tips of the shrinkage cracks.
Multiple shrinkage cracks show lower stress concentration at each crack tip. However, if
the geogrid stiffness is extremely large, additional shrinkage crack is not able to
significantly reduce the stress intensity factor at the crack tips of the existing shrinkage
cracks.
107
Table 7.2 Stress Intensity Factors of Shrinkage Cracks (Geogrid Stiffness = 400 kN/m)
5B
Label of
Studied Model
Number of
Cracks
Crack No.
7.4
7.5
7.6
7.7
7.8
K I ( MPa m )
Upper Crack
Lower Crack
Tip
Tip
0.153
0.116
0.137
0.108
ii
0.116
0.075
0.133
0.097
ii
0.097
0.062
iii
0.126
0.095
0.109
0.081
ii
0.095
0.057
iii
0.124
0.094
iv
0.149
0.092
0.107
0.079
ii
0.095
0.057
iii
0.123
0.094
iv
0.105
0.072
0.151
0.091
108
Table 7.3 Stress Intensity Factors of Shrinkage Cracks (Geogrid Stiffness = 800 kN/m)
56B
Label of
Studied Model
Number of
Cracks
Crack No.
7.4
7.5
7.6
7.7
7.8
K I ( MPa m )
Upper Crack
Lower Crack
Tip
Tip
0.140
0.116
0.127
0.108
ii
0.108
0.075
0.123
0.097
ii
0.090
0.062
iii
0.116
0.095
0.102
0.081
ii
0.089
0.057
iii
0.115
0.094
iv
0.149
0.092
0.100
0.079
ii
0.088
0.057
iii
0.114
0.094
iv
0.099
0.072
0.151
0.091
109
Table 7.4 Stress Intensity Factors of Shrinkage Cracks (Geogrid Stiffness = 1600 kN/m)
57B
Label of
Studied Model
Number of
Cracks
Crack No.
7.4
7.5
7.6
7.7
7.8
K I ( MPa m )
Upper Crack
Lower Crack
Tip
Tip
0.126
0.115
0.114
0.108
ii
0.099
0.075
0.110
0.097
ii
0.083
0.062
iii
0.105
0.095
0.093
0.081
ii
0.082
0.058
iii
0.103
0.094
iv
0.148
0.092
0.091
0.079
ii
0.082
0.057
iii
0.103
0.093
iv
0.092
0.072
0.151
0.092
110
Table 7.5 Stress Intensity Factors of Shrinkage Cracks (Geogrid Stiffness = 3200 kN/m)
58B
Label of
Studied Model
Number of
Cracks
Crack No.
7.4
7.5
7.6
7.7
7.8
K I ( MPa m )
Upper Crack
Lower Crack
Tip
Tip
0.111
0.115
0.101
0.107
ii
0.089
0.075
0.097
0.096
ii
0.076
0.062
iii
0.093
0.095
0.084
0.081
ii
0.075
0.058
iii
0.091
0.094
iv
0.147
0.092
0.082
0.079
ii
0.075
0.057
iii
0.091
0.093
iv
0.088
0.072
0.152
0.092
111
Table 7.6 Stress Intensity Factors of Shrinkage Cracks (Geogrid Stiffness = 6400 kN/m)
59B
Label of
Studied Model
Number of
Cracks
Crack No.
7.4
7.5
7.6
7.7
7.8
K I ( MPa m )
Upper Crack
Lower Crack
Tip
Tip
0.096
0.115
0.088
0.107
ii
0.081
0.075
0.085
0.096
ii
0.070
0.062
iii
0.082
0.094
0.075
0.081
ii
0.069
0.058
iii
0.080
0.094
iv
0.146
0.092
0.074
0.076
ii
0.069
0.057
iii
0.080
0.093
iv
0.085
0.072
0.152
0.092
112
Table 7.7 Stress Intensity Factors of Shrinkage Cracks (Geogrid Stiffness = 12800 kN/m)
60B
Label of
Studied Model
Number of
Cracks
Crack No.
7.4
7.5
7.6
7.7
7.8
K I ( MPa m )
Upper Crack
Lower Crack
Tip
Tip
0.083
0.114
0.077
0.106
ii
0.073
0.075
0.074
0.096
ii
0.065
0.062
iii
0.072
0.094
0.068
0.081
ii
0.064
0.058
iii
0.070
0.093
iv
0.146
0.092
0.067
0.078
ii
0.065
0.058
iii
0.071
0.093
iv
0.083
0.072
0.152
0.092
113
Stiffness = 0
Stiffness = 400
Stiffness = 800
Stiffness = 1600
Stiffness = 3200
Stiffness = 6400
Stiffness = 12800
0.200
0.180
0.160
0.140
0.120
0.100
0.080
0.060
0.040
0.020
7.4
7.5
7.6
7.7
7.8
0.000
i
ii
ii
iii
ii
iii
iv
ii
iii
iv
Crack No.
Figure 7.4 Mode I Stress Intensity Factor at Upper Crack Tip of Shrinkage Crack
82B
114
0.200
Model 7.4
0.180
Model 7.5
Model 7.6
0.160
Model 7.7
0.140
Model 7.8
0.120
0.100
0.080
0.060
0.040
0.020
0.000
0
2000
4000
6000
8000
10000
12000
Figure 7.5 Mode I Stress Intensity Factor at Upper Crack Tip of Crack No. i
83B
115
14000
0.200
Stiffness = 0
Stiffness = 400
Stiffness = 800
Stiffness = 1600
Stiffness = 3200
Stiffness = 6400
Stiffness = 12800
0.150
0.100
0.050
7.4
7.5
7.6
7.7
7.8
0.000
i
ii
ii
iii
ii
iii
iv
ii
iii
iv
Crack No.
Figure 7.6 Mode I Stress Intensity Factor at Lower Crack Tip of Shrinkage Crack
84B
116
8.1 CONCLUSIONS
40B
initial crack tips is evaluated in terms of the stress intensity factor. Compared to the
fracture toughness of the pavement materials, the Mode I stress intensity factor at the
initial shrinkage crack is large enough to drive the crack to propagate further. Finite
element modeling results show that the Mode I stress intensity factors of the shrinkage
crack increase with the increase of the crack length if the loading condition remains
unchanged. When the shrinkage crack propagates through the entire pavement structure,
a longitudinal crack will show at the pavement surface. This longitudinal crack should be
close to the pavement shoulder according to the location of the initial crack estimated by
theoretical calculation. This finding agrees with the field observations.
In practice, this type of longitudinal crack is controlled or minimized by a number
of methods. Geogrid reinforcement and lime treatment are the most effective methods,
but their application is limited to local experience. The second issue addressed in this
dissertation is the investigation of the mechanism of geogrid reinforcement and lime
treatment. Based on the development process of the longitudinal cracks due to shrinking
expansive subgrade, two approaches can be used to control the shrinkage crack
propagation: 1) reducing the stress concentration at the shrinkage crack tips, and 2)
increasing the fracture toughness of the pavement layers.
The geogrid is able to reduce the stress concentration at the crack tip if it is placed
at the bottom of the base (directly on top of the subgrade). Since the initial shrinkage
crack develops close to the interface of the base and subgrade, the shrinkage crack is at
its early stage when it reaches the geogrid. At this stage, the shrinkage crack has a
relatively low stress intensity factor at its upper crack tip. When the shrinkage crack goes
through the geogrid apertures, the geogrid has to deform with the crack opening. The
deformation of the geogrid results in forces that are directly applied to the crack surfaces,
tending to close the crack. These forces significantly reduce the stress intensity factor at
118
the upper tip of the shrinkage crack. A geogrid with a higher stiffness is able to apply
larger forces to the crack surfaces at low deformation. More reduction in stress intensity
factor is then produced at the upper tip of the shrinkage crack. As a result, a geogrid with
higher stiffness is more efficient for the control of the shrinkage crack.
The lime treatment can improve the mechanical properties of the expansive soil.
The lime-treated soil has lower plasticity index, higher tensile strength and higher
fracture toughness. Finite element modeling results show that the tensile stress
developing in the lime-treated layer is less than the tensile strength of the lime-stabilized
soil. Therefore, shrinkage cracks are less likely to initiate in the lime-stabilized layer. The
possible location of the shrinkage crack initiation is in the untreated soil close to the
bottom of the lime-treated layer, where tensile shrinkage stress exceeds the tensile
strength of the untreated natural soil. When the initial crack propagates into the limetreated soil, the magnitude of the stress intensity factor at the upper crack tip is in the
same order of the estimated fracture toughness of the lime-treated soil. The shrinkage
crack is less likely to develop through the lime-treated soil with increased fracture
toughness. If multiple shrinkage cracks are present in the untreated soil, the strain energy
release rate at each crack tip is reduced because the strain energy is released by multiple
cracks. The possibility of crack propagation is then further decreased.
The combination of geogrid reinforcement and lime treatment offers the most
reduction in the stress concentration at the crack tips. The best place to install the geogrid
is at the interface of the lime-treated layer and the untreated natural soil. At this location,
the geogrid has the most efficient reinforcement. The geogrid with higher stiffness
provides more benefit. If using a geogrid with high stiffness, the Mode I stress intensity
factor may be reduced to a certain level that is lower than the fracture toughness of the
pavement material. The development of multiple shrinkage cracks can further reduce the
119
stress concentration at the crack tips. However, when the geogrid stiffness is extremely
high, the effect of additional shrinkage cracks is not significant.
8.2 RECOMMENDATIONS FOR FURTHER RESEARCH
41B
As much as this study has attempted to bridge the gap between current
engineering practice and the theoretical analysis for the longitudinal crack problem due to
expansive subgrade, it also raises the need for further investigation on a number of
research problems. These problems must ultimately be resolved if the analytical
methodologies proposed in the study are to become practical for pavement design. The
most significant research topics include:
1. The fracture properties of pavement materials used on this study were assumed
based on the limited literature. The fracture toughness of some pavement
materials has not been addressed in the available literature. The lack of exact data
about material fracture toughness makes it difficult to accurately determine
whether or not the shrinkage crack is stable in the subgrade. An extensive
laboratory experiment should be developed and conducted to collect detailed data
about the fracture toughness of asphalt concrete, granular material, lime-treated
soil and natural soil. The use of tested instead of assumed fracture properties will
be beneficial for more accurately judging the dynamic status of the shrinkage
crack in the pavement structure.
2. This research found that the geogrid stiffness had a significant effect on the stress
concentration at the crack tips. A variety of stiffness values were used for the
geogrid in the finite element modeling. However, the mechanism of the geogrid
reinforcement in this particular problem is different from the traditional
application of geogrid-reinforced pavements. The forces that the geogrid applies
to the crack surfaces are directly from the shear deformation (the deformation in
120
the direction perpendicular to the axial direction of the ribs) of the geogrid ribs
instead of the deformation in the axial direction. The geogrid stiffness values used
in the proposed model should not be tested by the standard testing method for
determining tensile properties of geogrid (ASTM D 6637-01). Instead, an
experiment should be developed to simulate the actual loading condition of the
geogrid. In this experiment, the forces should be placed at the geogrid ribs
perpendicular to the rib axial direction.
3. Besides geogrid reinforcement and lime treatment, geotextile materials without
apertures may also be effective for the control of shrinkage cracks owing to a
different mechanism. Since the geotextile does not have apertures, it is difficult
for the shrinkage crack to propagate through the geotextile. If the geotextile is not
full bonded to the pavement materials but has a slippery surface, interface cracks
(Mode II cracks) may develop at the interface of the geotextile and the subgrade
soil. The strain energy is then released by the Mode II cracks at the interface. The
Mode I shrinkage crack may stop developing at the interface. For this situation,
the mixed mode cracking theory in layered materials (Hutchinson and Suo, 1992)
can be used to examine the interface crack propagation. The use of geotextiles
may be a good alternative to properly control longitudinal cracks in response to
moisture level changes in expansive subgrade.
4. The proposed research focused on the benefits of treatment methods from the
mechanical point of view. The financial cost, the construction complexity, and the
construction quality control should also be taken into consideration. It is desirable
to conduct cost-effective analysis on different treatment methods. These analyses
will provide additional support for the preference for any one of the treatment
methods.
121
Bibliography
10B
122
124
125
126
Vita
1B
Rong Luo was born in Yibin, Sichuan Province, China on January 12, 1979, the
daughter of Xiuzhong Luo and Xingwen Chen. After graduating from Yibin No. 1 High
School in 1997, she entered Chongqing Jiaotong University in Chongqing, China. She
received a Bachelors degree in July 2001 and a Masters degree in June 2003, both in
Civil Engineering, from Chongqing Jiaotong University. She joined the Ph.D. Program in
Transportation Engineering at The University of Texas at Austin in August 2003. Her
research has been focused on mechanism of pavement cracking development, application
of geosynthetics in pavement structure, interaction between truck tire and pavement
surface, and nondestructive pavement testing. She was awarded the 2005 Wanda J.
Schafer Graduate Scholarship and the 2007 Helene M. Overly Memorial Scholarship by
the Womens Transportation Seminar (WTS). She also received the 2007 International
Road Federation (IRF) Executive Leadership Fellowship and was named an IRF
Executive Leadership Fellow. She has been appointed as a Young Member of the
Transportation Research Board (TRB) Geosynthetics Committee and International
Activities Committee.
127