Академический Документы
Профессиональный Документы
Культура Документы
133
for the design period. However this method has a limitation on determining subgrade
deviator stress, since it is based upon a limited number of test results on high plasticity
soils that generally represent weak subgrade condition.
Rose et al (2006, 2010) presented a performance-based layered elastic railway
trackbed structural design and analysis using KENTRACK program, which was
initially developed to analyze traditional all-granular layered trackbeds and asphalt
layered trackbeds. The current version of this program includes option to
accommodate a composite track substructure dealing with a combination of granular
and asphalt layers. Similar to the M-E PDG program, it predicts the performance lives
with respect to the subgrade vertical stress and fatigue cracking of asphalt subballast
layer to determine design layer thickness. However it also exhibits limitations on
adopting the performance model that has originally been developed from pavement
engineering.
The reinforcement of substructure using geosynthetics or chemical
stabilization has been found to be effective for increasing load bearing capacity and
extending service life. Leshchinksy and Ling (2013) conducted numerical modeling
of the behavior of ballasted railway with geocell confinement. The study revealed that
three dimensional confinements of geocell improved load bearing capacity
significantly when used on a wide range of subgrade stiffness, when using weaker
ballast.
Scullion et al (2008) conducted a study with the Portland Cement Association
(PCA) to review the proposed models for soil-cement base and cement modified soils
incorporated into the M-E PDG program. They proposed a new test to measure key
material properties such as resilient modulus and modulus of rupture for such
stabilized materials. For the design purpose, the performance prediction models were
calibrated and built into two software packages developed in this study.
In this study, the effectiveness of geo-reinforcement techniques for track
substructure enhancement using cement treated subballast, asphalt subballast, and
application of geosynthetics is evaluated via a limited numerical analysis in terms of
thickness design that related to material savings.
NUMERICAL ANALYSIS SETUP
In this study, the author used a finite element program PLAXIS 2D (Brinkgreve et al.
2006) under plane strain condition of 15 node elements for the parametric study. Due
to symmetry, only one half of the track section was considered in the numerical model
as shown in Fig. 1. To accommodate geo-reinforcement modeling, the following
substructures were considered:
Geogrid placed at the one-third depth from the bottom of the ballast layer
Asphalt concrete subballast
Cement treated subballast
Oh (2013) found that the placement of geogrid at the one-third depth from the bottom
of the ballast layer exhibited the most effectiveness in controlling substructure
deformation.
134
135
With respect to the boundary condition, both sides were set to move vertically
while the bottom was fixed to prevent any movement. For the traffic loading, an
equivalent dynamic wheel load (Pdl) for a given static wheel load (Psl) was computed
as per the American Railway Engineering Association (AREA 1996) approach and is
given by:
0.0052V
Pdl = 1 +
Psl
D
(1)
Where V and D are train speed (in km/h) and wheel diameter (in meter), respectively.
For the parametric study, a representative vertical load was applied to produce
acceptable ballast sleeper interface stress within 450 kPa in a control ballasted track
without geo-reinforcement. (US Army of Corps Engineers, 2000). Based on equation
(1), an equivalent dynamic wheel load of 260 kN was applied with given vehicle speed
equal to 200 km/hour, 1 meter of wheel diameter, and 125 kN of static wheel load.
With regard to geogrid modeling, a geogrid element provided by PLAXIS was
employed along with interface elements that are connected with adjacent track
substructure layers. The only property in a geogrid data set is the elastic axial stiffness,
EA, in terms of force per unit width. Based on the study conducted by Oh (2013), 500
kN/m was used in this study.
Sleeper
Ballast
Geogrid
Sub-ballast
Subgrade
136
hardening soil model was applied for ballast layer as given in equation (2).
E50 = E
ref
50
c cot + 3
ref
c cot + p
(2)
Where E50 is a confining stress (3) dependent secant modulus at 50% strength for
primary loading, pref is a reference confining pressure, E50ref is a reference modulus for
primary loading corresponding to the reference confining pressure pref, c is a cohesion,
is a friction angle, and m is a factor ranging from 0.5 to 1.0. For the granular
subballast and subgrade layer, the Mohr-Coulomb model was employed to capture the
plastic points that indicate the stress states exceed the yield surface. Asphalt subballast
layer was modeled as linear elastic material even though asphalt material is well
known as viscoelastic material susceptible to the change of temperature and loading
frequency.
Cement treated subballast layer was also modeled as linear elastic material.
According to the Scullion et al (2008), the typical modulus of cement treated base
layer ranges from 344.5 to 13780 MPa, while the cement modified soil has 172 to 689
MPa of resilient modulus. Table 1 summarizes material properties used in parametric
study.
Table 1. Material Properties used in the Finite Element Analysis
Property
Sleeper
Ballast
Sub
ballast
Model
Thickness
(m)
(kN/m3)
E (MPa)
E50ref (MPa)
Eoedref (MPa)
Eurref (MPa)
ur
EA (kN/m)
C(kN/m2)
(degree)
(degree)
Pref (kN/m2)
m
K0nc
Rf
Elastic
HS
MC
AC
Subballast
Elastic
0.2
0.3
0.2
24
10340
0.3
-
15.3
70
67
210
0.1
45
0.0
100
0.5
0.3
0.9
19
100
0.35
1.0
30
0.0
-
CementSub-ballast
Sub
grade
Elastic
MC
0.1
0.1
3.5
23
1500
0.38
-
17
1000
0.25
-
17
50
0.4
5.0
30.0
0.0
-
137
Eoedref: tangent stiffness for primary oedometer loading, Eurref: triaxial unloading/reloading
stiffness, ur: Poissons ratio for loading conditions, = effective friction angle, =
dilatancy angle, K0nc =coefficient of earth pressure at rest for normal consolidation,
Rf=failure ratio.
PERFORMACE CRITERIA
It is crucial to establish performance criteria to conduct mechanistic thickness design
for pavement or railroad. There are various performance criteria for M-E PDG
depending on material and pavement types. In this study, the following performance
criteria were adopted despite of limitation in applying these for the evaluation of track
substructure thickness design.
Where
Na
t
Ea
Nd
c
Es
(3)
(4)
Equations (3) and (4) are employed to assess the performance of asphalt subballast and
subgrade layer respectively, which is incorporated into the KENTRACK program.
With regard to the cement treated subballast layer, the following fatigue relationship is
employed in this study as follows:
0.972 c1 t
M
rup
log N f =
0.0825 * c 2
(5)
Where
Nf
= number of repetitions to fatigue cracking of the stabilized layer;
t
= maximum traffic induced tensile stress at the bottom of the stabilized layer
(psi);
Mrup = 28 day modulus of rupture (Flexural Strength) (psi); and
c1, c2 = field calibration factors.
Scullion et al (2008) used equations (5) in the developed program. A design limit for
accumulative fatigue damage is set to be 25% of the total fatigue life. The study
138
Control
Geogrid
AC
Subballast
Cement
Subballast
321.6
288.9
325.6
326
124.9
11
113.4
16
130.4
0.10
3.75
128.9
1000
All cases were found to meet the criterion of ballast-sleeper interface stress by
exhibiting vertical stress level less than 448 kPa. However, it should be noted that the
interface stress was slightly reduced with geogrid reinforcement compared to the
control case.
The number of repetitions to failure was computed using equations (3) to (5)
for each case. For the control and geogrid reinforced sections, the performance
criterion based on subgrade stress was only considered using equation (4). With
respect to asphalt concrete sub-ballasted track, both criteria were applied using
equations (3) and (4). The performance of cement treated sub-ballasted track structure
was evaluated using equations (5). The allowable number of repetitions was assumed
as 10000 to estimate damage factor, which is computed by dividing allowable number
of repetitions by the number of repetitions to the failure. If damage factor is less than
a value of one, then the structure can be expected to exceed its design life, otherwise
the structure is expected to fail prematurely. However the value is much less than one,
the structure is probably designed too conservatively. Table 3 summarizes the
analysis results.
Table 3. Comparison of Service Life based on Performance Criteria
Damage Factor
Control
Geogrid
AC
Cement
Subgrade failure
Asphalt failure
Cement treated failure
1.5
-
1.1
-
Subballast
1.8
6.0
-
139
Subballast
1.7
6.9
The results presented in Table 3 indicate that all cases showed a slight degree of
damage with respect to subgrade failure. However it should be noted that when the
substructure was reinforced with asphalt concrete and cement stabilization, the
damage factor slightly increased due to increased subgrade compressive stress, which
was similarly found from a previous study conducted by Park et al (2012). The author
is of the opinion that the presence of high stiffness layer between ballast and subgrade
might attribute to higher confinement thorough subgrade.
The damage factor in terms of asphalt failure and cement treated failure
yielded significantly high number, requiring alternative design. An attempt was made
to modify the current design in order to meet the performance criteria. The following
findings were obtained.
Increased 5 cm of AC sub-ballast and reduced 5 cm of ballast reduces AC
tensile strain by 30 percent and reduces subgrade compressive stress by 6
percent. In this case, since the damage factors tend to be closer one, this
design seems to be more appropriate.
Increased 5 cm of cement treated sub-ballast and reduced 5 cm of ballast
reduces cement treated tensile strain by 7 percent and reduces subgrade
compressive stress by 4 percent. In this case, the damage factors with respect
to the cement treated failure significantly reduced up to 1.4, even though the
degree of reduction of tensile stress was not that substantial. Therefore, this
design seems to be more appropriate with regard to the performance criteria.
CONCLUSIONS
140
141