You are on page 1of 11

Construction and Building Materials 362 (2023) 129654

Contents lists available at ScienceDirect

Construction and Building Materials


journal homepage: www.elsevier.com/locate/conbuildmat

Behavior of sand-contaminated ballast reinforced with geogrid under


cyclic loading
Javad Sadeghi *, Ali Reza Tolou Kian, Ali Khanmoradi, Mohammad Chopani
School of Railway Engineering, Iran University of Science and Technology, Tehran, Iran

A R T I C L E I N F O A B S T R A C T

Keywords: A ballasted railway track is contaminated when passes through sandy areas. Due to sand penetration into the
Ballast ballast voids, mechanical behavior of tracks deteriorates. Despite the dynamic nature of the load in the railway
Geogrid system, mechanical behavior of ballast reinforced with geogrid and contaminated with sand under cyclic loading
Ballast box test
have not been investigated. To study the influence of geogrid on improvement of cyclic mechanical properties of
Sand contamination
Railway track
contaminated ballast, extensive laboratory tests were performed in this research. To this end, a ballast box
Settlement machine was used in order to model in-situ cyclic loads. The tests were carried out on ballast samples
contaminated with various sand dosages and reinforced with different geogrids. The results indicate that sand
materials make ballast aggregates move and slide easier, causing more ballast settlement. It was shown that by
geogrid reinforcement, settlement decreases (up to 75%), degradation (breakage) of ballast reduces (up to 23%),
and its stiffness increases (up to 50%). An innovative model was developed which takes into account sand
content and geogrid reinforcement that provides the settlement of ballast. An effective method of geogrid
reinforcement for obtaining the best performance of ballast under cyclic loading was derived.

1. Introduction results derived from the box tests indicates realistic ballast responses
when compared with test results obtained from track-beds. Tutumluer
Sand contamination of ballast, as reported in the USA, Asia, Middle et al. [54] investigated ballast settlements through simulation of ballast
East and North Africa, makes deterioration of ballast mechanical box test under cyclic loading using discrete element method (DEM).
behavior, and consequently causes halts in railway operation [5,51–52]. They showed that a constant load with low frequency causes consider­
Two views of railway contamination in eastern railway network of Iran able plastic deformations in ballast. Sol-Sanchez et al. [49] studied the
are presented in Fig. 1. It has been shown that as ballast is contaminated, influence of crumb rubber in ballast using ballast box test. They indi­
sand penetrates into the ballast voids and the ballast cannot fulfill its cated that addition of 10% of crumb rubber (by volume) to the ballast
functions in the track system [44–45,12,3,57,53,37]. mass reduces ballast degradation. Choi et al. [7] developed a ballast
A survey of the literature elucidates that the axes of principal stresses settlement model using the results of ballast box tests. Gao and his
in railway track layers rotate in the plane parallel to the direction of colleagues [19] developed a ballast box model using results of a DEM
train traffic. The rotation of principal stress axes is in harmony with the analysis. Their results reveal that ballast settlement reduces with the
changes of longitudinal and shear stresses in the direction of train traffic. increase in the longitudinal force applied to the ballast.
When the train moves, the magnitude of principal stresses varies In regard to track contamination condition, some experimental and
[34,18]. Due to the dynamic nature of the railway loads; one of the most numerical studies were carried out on contaminated ballast. In these
reliable studies of ballast materials is to investigate ballast behavior studies, contamination materials were considered as small crushed
under cyclic loads. Extensive laboratory research and several numerical ballast grains, coal dust and clay [28,33,40]. Results of laboratory tests
studies have been performed on ballast cyclic behavior. McDowell et al. conducted on contaminated ballast elucidate that when ballast gets
[35] have conducted some box tests to study the influences of train loads contaminated; stiffness and strength of ballast decrease gradually
and tamping operation on the behavior of ballast. They proved that [28,33,40]. Sadeghi et al. [47] investigated effects of different particle

* Corresponding author.
E-mail addresses: javad_sadeghi@iust.ac.ir (J. Sadeghi), ar_tkian@iust.ac.ir (A.R. Tolou Kian), ali_khanmoradi@alumni.iust.ac.ir (A. Khanmoradi), mohammad_
chopani@alumni.iust.ac.ir (M. Chopani).

https://doi.org/10.1016/j.conbuildmat.2022.129654
Received 21 August 2022; Received in revised form 9 October 2022; Accepted 1 November 2022
Available online 15 November 2022
0950-0618/© 2022 Elsevier Ltd. All rights reserved.
J. Sadeghi et al. Construction and Building Materials 362 (2023) 129654

size distribution of ballast on ballast behavior under cyclic loading in response of ballast contaminated with sands was examined through
sand contamination condition. They showed that in a high contamina­ laboratory tests in this research. For this purpose, extensive cyclic tests
tion level; ballasts mechanical properties were significantly worsened were carried out on ballast samples contaminated with various dosages
regardless of their gradations [47]. of sand and reinforced with geogrid. The samples were tested using a
For reduction of adverse effects of contamination and to improve ballast box machine. Geogrids with different aperture sizes installed at
contaminated ballast mechanical behavior, application of geogrid- various positions within the ballast were examined. The analysis of the
reinforcement has been studied in the previous researches. results led to obtain optimum application of geogrid (regarding aperture
[10,42,41,29,6,31]. Indraratna et al. [23] showed that geogrid aug­ size and position of geogrid) in sand-contaminated ballast. Also, a new
ments the shear strength and friction angle of ballast. They also showed settlement prediction model was extended which considers simulta­
that geogrid makes reduction in the vertical displacement of the ballast neously geogrid presence and sand content. Using the results obtained, a
sample. Esmaeili et al. [15] investigated the influence of geogrid on the practical solution for improving ballast mechanical behavior in railway
lateral resistance of a ballasted track through laboratory and field tests. tracks which pass through desert/sandy areas was suggested.
Brown et al. [4] investigated cyclic settlement of clean ballast reinforced
with geogrid using a composite element test machine. They indicated 2. Laboratory investigation
that geogrid rib cross-sectional shape and stiffness are parameters that
have an influential effect on the effectiveness of reinforcement. [23,27] In order to examine the responses of ballast subjected to repeated
surveyed the effect of geogrid opening size on shear strength of ballast, train loads, thorough laboratory tests were conducted. In the tests, cyclic
they performed extensive laboratory tests including direct shear tests on vertical loads were applied on ballast samples using a ballast box
ballast reinforced with geogrid and contaminated with various dosages apparatus. Unreinforced (i.e., without geogrid) and ballast samples
of clay. [23,27] developed a model to estimate the shear strength of reinforced with geogrid were made with various contents of sand. Based
geogrid-reinforced ballast. Liu et al. [33] studied the effect of geogrid on Selig and Waters’ suggestion (1994) the percentage of contamination
reinforcement on horizontal displacement of clean ballast aggregate was taken as an index, representing contamination level. The percentage
using ballast box tests. Their results indicate that horizontal displace­ of contamination is defined as the ratio of the weight of contaminant
ment and rotation are influential modes of ballast grain movement material which passes through #4 sieve (or 9.5 mm sieve) to the total
under cyclic loading. Yu et al. [56] performed experimental research weight of the sample [48]. Ballast samples with percentage of sand
into the effectiveness of geogrids on confining ballast materials. They contamination of 0% (i.e.; clean ballast), 7%, 14%, 24%, and 36% were
showed that geogrid have the most benefits when the track subgrade is made for the tests. That is, 11.3 kg, 24.6 kg, 47.7 kg and 85.0 kg sand
soft. Sadeghi et al. [46] carried out plate load tests and large-scale direct materials were added to ballast to prepare samples with 7%, 14%, 24%,
shear tests on geogrid-reinforced ballast samples with various dosages of and 36% of sand contamination. Ballast and sand material were kept dry
sand. They showed that geogrid reinforcement is effective when sand in the all samples. Three different geogrids that had square opening were
contamination is not noticeable. Sadeghi and his colleagues [46] did not used in the tests. Aperture sizes of geogrids are 24 mm, 34 mm and 46
perform cyclic test on ballast contaminated with sand and therefore, mm. Technical features of the geogrids are indicated in Table 1. Ac­
their research does not cover ballast behavior under cyclic loading. In cording to Table 1, these geogrids have approximately the same strength
other words, settlement under cyclic loads, damping, breakage, and and the difference is only in the size of their openings. The reinforce­
stiffness of sand contaminated ballast reinforced with geogrid have not ment mechanism of geogrid in granular materials originates from
been investigated. confinement of the particles which leads to reduction in lateral
Although cyclic behavior of ballast contaminated with coal dust, clay spreading of the particles [11]. To achieve the best confinement of the
and small crushed stones have been studied, the application of the re­ particles; an effective interaction between the particles and geogrid
sults in sand contaminated ballast is questionable because geotechnical aperture should be made. That is, the size of the aperture should be in
properties of sand are significantly different from coal dust, clay and harmony with the diameter of the particles to trap and confine the
small crushed stones [53]. particles effectively. Based on the particle size distribution of ballast
A survey of the literature elucidates that application and efficacy of stones and d50 of ballast particles (d50 = 32 mm) indicated in Fig. 2,
geogrid in improvement of sand-contaminated ballast behavior have three types of geogrids were selected. The first one (24 × 24 geogrid) has
been studied in static state of the load [46]. However; application of opening size less than d50 of ballast particles; the second one (46 × 46
geogrid in ballast contaminated with sand under cyclic loading has not geogrid) has opening size greater than d50 of ballast, and the third one
been investigated. To fill this gap, the effect of geogrid on the cyclic (34 × 34 geogrid) has opening size close to d50 of ballast. According to

Fig. 1. A ballasted track in Iran railway system in sand contamination condition.

2
J. Sadeghi et al. Construction and Building Materials 362 (2023) 129654

Table 1 Particular designation was attributed to each sample. In this research,


Mechanical specification of geogrid used in the current paper (adapted from name of reinforced samples is “Ssand contamination percentage-Ggeogrid type-Hpo­
[58] and [38]. sition of geogrid”. “sand contamination percentage” corresponds to each of 0,
Specification 46 × 46 geogrid 34 × 34 24 × 24 7, 14, 24, and 36. “type of geogrid” corresponds to any of G24, G34 and
geogrid geogrid G46. “position of geogrid” corresponds to 10 or 20 which matches up the
Geogrid type Polypropylen Polyester Polyester position of geogrid in centimeters from the base of the ballast. For
welded geogrid welded geogrid extruded instanse, “S14-G46-H10” represents a sample with 14% sand content
geogrid that is reinforced with G46 geogrid that is installed within ballast
Opening size (mm × 46 × 46 34 × 34 24 × 24
located 10 cm from the the ballast bottom. Name of an unreinforced
mm), md × cmd*
Mass in unit area (g/ 300 320 380 sample is “Ssand percentage-WOGeo”; WOGeo indicates that the sample has
m2) no reinforcement.
Greatest tensile More than 30/ More than 30/ More than 30/
strength (kN/m), Greater than 30 Greater than 30 Greater than 30
2.1. Materials in tests
md/cmd*
Elongation at nominal Smaller than 8/ Smaller than 7/ Smaller than 7/
strength (%), md/ Smaller than 8 Smaller than 7 Smaller than 7 Fresh andesite aggregates obtained from Shahriar quarry was used
cmd* for ballast materials. Particle size distribution of ballast was selected in
Tensile strength at 2% 10.5/10.5 13.5/13.5 13.5/13.5
accordance with Iran Railway [8]. Also, windblown sand as a contam­
elongation (kN/m),
md/cmd*
ination material was sampled from Iran sandy parts (Tabas region)
Tensile strength at 5% 24/24 24/24 24/24 where railway tracks pass. It means that there is no difference between
elongation (kN/m), particle size distribution of sand materials used in this study and those
md/cmd* exist in deserts and sandy areas where the actual train is running. The
*cmd and md stand for machine cross direction and machine longitudinal di­ properties and size distribution of sand material and ballast are illus­
rection respectively. trated in Table 2 and Fig. 2, respectively. The aperture sizes of geogrids
used in this study (i.e., 24 mm, 34 mm and 46 mm) are selected in order
the information provided by manufacturers, the stiffness values of the to cover the aperture sizes which are equal to, greater, or smaller than
geogrids used in this study are almost the same. According to the liter­ 50% of the ballast aggregates. In other words, a parametric experimental
ature, [42]; the stiffness of the geogrid affects its performance. That is, a study was done to derive the optimum aperture size of geogrid.
geogrid with higher stiffness leads to less ballast permanent deforma­
tion. The tests were conducted in two stages; in the first stage, geogrid 2.2. Test set up and method of testing
sheet was installed within ballast material at a height of 10 cm from the
bottom of the box. In the second stage, geogrid sheet was installed To study the effects of geogrid reinforcement on the cyclic settlement
within ballast at a height of 20 cm above the base of the box. These of ballast, a ballast box machine was used. This machine can simulate
distances were selected to satisfy two criteria (i) geogrid should be away cyclic loads applied on ballast in railway tracks. The machine comprises
from tamping reaching zone (ii) to obtain the position of geogrid which
has the most effects on improvement of ballast properties. Table 2
In this laboratory investigation, 35 samples were prepared. There Sand and ballast size properties.
were five unreinforced samples, ten samples reinforced with G24 (i.e., Material Specific gravity dMax mm dMin mm d50 mm
geogrid which has opening size of 24 mm), ten samples reinforced with
Andesite Ballast 2.78 60 9.4 32
G34 (i.e., geogrid which has opening size of 34 mm), and ten samples Sand 2.68 1.0 0.075 0.23
reinforced with G46 (i.e., geogrid which has opening size of 46 mm).

Fig. 2. Sand and ballast material that are used in this study and their grain size distributions.

3
J. Sadeghi et al. Construction and Building Materials 362 (2023) 129654

of a rigid frame, a hydraulic actuator, a load cell, a solid box, a linear


variable displacement transducer (LVDT), a data logger and a computer
system. The overall view of the machine is presented in Fig. 3. The load
was applied on the sample by the hydraulic actuator. The load was
adjustable and it was under control during the tests. The load cell was
connected to the actuator to measure forces applied to the samples. The
load cell and LVDT were connected to the data logger which records all
the data from these sensors. The solid box of the machine has inside
depth, width and length of 450 mm, 300 mm and 700 mm, respectively
as illustrated in Fig. 3c. Dimension of the box in this study is selected
such that it simulates a part of ballast under the rail seat as depicted in
Fig. 4. According to Lim’s studies, results of ballast settlement and
aggregate breakage derived from the box test are in good agreement
with results obtained from in-situ tests [32]. There are some experi­
mental researches into the response of ballast materials under cyclic
loading using ballast box test. The dimension of the box and dMax of
ballast stones in the available researches are indicated in Table 3.
The settlement of samples is measured by the LVDT which was
connected to loading plate as shown in Fig. 3b. The plate was placed on
top surface of the ballast sample. The square loading plate has length of
Fig. 4. View of ballasted railway track presenting section simulated by the
220 mm. Accuracy, reliability and repeatability of the results obtained
ballast box test.
from the machine were evaluated and confirmed by [13–14].
To fill the solid box, ballast materials were placed in three layers.
there is small difference between the ballast density of the test and that
Each layer was compacted by use of a tamper. By computing the ratio of
one measured in railway track field. In addition; density of ballast
the ballast weight to the volume of the box filled by ballast, the density
samples in the box tests ranges from 1500 to 1700 kg/m3 [47,16,17]. To
of the clean ballast sample was derived in dry condition. It was 1600 kg/
make ballast contaminated in the box, sand was dispersed on the top
m3. A comparison was made between density of ballast in the test and
surface of each ballast layer. This method was adapted in accordance
that obtained from in-situ tests [53]. The comparison indicates that
with the findings in the available literature [43,24,20]. For each layer of

Fig. 3. Ballast box machine and solid box.

4
J. Sadeghi et al. Construction and Building Materials 362 (2023) 129654

Table 3 versus the number of load cycles is illustrated in Fig. 6a to 6d. The effect
Dimension of the box and dMax of ballast particles in the literature. of geogrid on settlement of clean ballast is shown in Fig. 6a. The effect of
Researcher Box size (mm) dMax (mm) geogrid opening dimension on settlement of ballast samples contami­
nated with sand is illustrated in Fig. 6a and 6c when the geogrid is
Selig and Waters [48] 300(W) × 610(L) × 460(D) Unknown
Lim [32] 300(W) × 700(L) × 450(D) 50 placed 10 cm and 20 cm from bottom of the box, respectively. The effect
McDowell et al. [35] 300(W) × 700(L) × 450(D) 50 of geogrid location on the settlement of samples is presented in Fig. 6b
McDowell and Stickley [36] 300(W) × 700(L) × 450(D) 63 and 6d when geogrid opening dimensions are 34 mm and 46 mm,
Sol-Sanchez et al. [49] 200(W) × 460(L) × 300(D) 63 respectively.
[13–14] 300(W) 700(L) × 450(D) 70
As illustrated in Fig. 6, settlement of all ballast samples stayed un­
×
[16–17] 300(W) × 700(L) × 450(D) 63
Mehran Khoshoei et al. [37] 300(W) × 700(L) × 450(D) 63 changed when the number of loading cycles becomes one hundred
Sadeghi et al. [47] 300(W) × 700(L) × 450(D) 63.5 thousand. Because the method of compaction was the same for all
W, L and D denote to width, length and depth.
samples, differences in settlements of the samples originate from ballast-
geogrid interlocking and position of geogrid. That is, when ballast ag­
gregates are better confined with geogrid (due to the opening size and
ballast; weight and depth of the ballast layer were measured; as a result,
position of geogrid) settlement of sample is decreased. The trend of
preparation of the sample in the box was under control.
settlement results indicates that the settlement after 100,000 cycle in­
For samples reinforced with geogrid, two groups of tests were done.
creases negligibly. According to results presented in Fig. 6, as sand
First, the geogrid was placed at a height of 10 cm from the bottom of the
percentages increase, settlement of samples is augmented in both rein­
box (as shown in Fig. 5b). For the second group, the geogrid was located
forced and unreinforced samples. This augmentation in settlement is
at a height of 20 cm from the bottom of the box. For the samples rein­
obtained because ballast grains can move, rotate and reposition in an
forced with geogrid, the tests were performed by use of G24, G34 and
easier manner with presence of sand material when compared to cases in
G46 geogrids, independently. In order to begin the test, cyclic loads were
which there are no small sand particles. In other words, because
applied to the ballast box using the hydraulic actuator. The frequency of
gradation sizes of sand are much less than those of ballast (as shown in
cyclic loads and their magnitude were set to 3 (Hz) and 30 (kN),
Fig. 2 and Table 1), there are more movement, sliding and rearrange­
respectively. The frequency of loading corresponds to a train running at
ment of ballast stones due to sand infiltration into the ballast voids.
a speed of 130 (km/h) while the distance between the centers of bogies
Results illustrated in Fig. 6a and 6d show that unreinforced samples
of a rolling stock is 12 m. The magnitude of cyclic loads matches up train
have greater settlement when compared to the reinforced ones under
axle load of 120 kN which is close to the axle load of the Alfa-Pendular
cyclic loads. Ballast settlements for S0-WOGeo, S0-G24-H10, S0-G34-
passenger train [9]. That is; the under wheel sleeper takes only 50% of
H10 and S0-G46-H10 samples (as indicated in Fig. 6a) were obtained.
the wheel load and the other 50% of the load is sustained with the
They were 4 (mm), 3.3 (mm), 1(mm) and 2.8 (mm) respectively. It in­
adjacent sleepers according to [55]. Also, the wheel load is equal to the
dicates that geogrid decreases ballast settlement up to 75% in clean
half of the axle load; therefore, the magnitude of cyclic load test was
condition when G34 geogrid is used at the height of 10 (cm) from the
considered 30 (kN). Taking into account the extent of the load applied
ballast base. The settlement reduction happens because geogrid limits
on the ballast, the maximum stress imposed on the sample was less than
and confines transverse displacement of ballast stones. This makes
750 kPa (the utmost limit suggested by [1].
ballast samples have less settlement. By comparison of the settlements of
In each test, one hundred thousand load cycles were applied on the
the reinforced samples (as illustrated in Fig. 6b and 6d), it can be seen
sample. Vertical settlement of the samples was recorded automatically
that samples with geogrid placed at the height of 10 cm (from the base of
in each load cycle by the data logger. To make sure the tests were
the box) have less settlement. Settlements of all the ballast samples at
repeatable, the method of testing was held constant throughout labo­
the end of the test (i.e., settlement at 100,000 load cycle) against dos­
ratory tests. To make least errors in the tests, any test was repeated two
ages of sand contamination are summarized in Fig. 7.
times and the results obtained were averaged.
According to Fig. 7, regardless of geogrid position, samples with G34
geogrid show less settlement when compared with the other reinforced
3. Laboratory results and discussions
samples. The reason originates from a better interlocking of ballast ag­
gregates with grids of G34 geogrid. The mechanism of reduction in
In this section, settlement and degradation (breakage) of the samples
ballast settlement (made by geogrid) is the interlocking between grid
were obtained and analyzed. Also, stiffness and damping of the samples
and ballast grains. 34 × 34 geogrid has contact and interaction with
were derived and discussed.
more number of ballast aggregates than the other geogrids used in these
tests. Comparing the results presented in Fig. 6a with those in Fig. 6d
3.1. Settlement under cyclic loads reveals that as sand content is high, the differences among the settle­
ments of the samples become smaller. That is, by increases of sand
The settlement of ballast samples without geogrid and with geogrid

Fig. 5. Ballast placement inside the box.

5
J. Sadeghi et al. Construction and Building Materials 362 (2023) 129654

Fig. 6. Ballast samples settlement under cyclic load.

3.2. Development of settlement prediction model for geogrid-reinforced


ballast

A survey of the literature elucidates that prediction models of ballast


settlement have been presented for clean ballasts and ballast contami­
nated with sand, clay and coal dust [25,26]. However, settlement of
ballast reinforced with geogrid in sand contamination has not been
investigated. The results obtained in this research were used to derive a
prediction model for sand-contaminated ballast reinforced with geogrid.
Different mathematical models were fitted to the settlement versus
number of load cycles. It was found that the logarithmic model is the
best fitted model for the development of the prediction expression. The
general mathematical equation of the model is as follows:
S = A × ln(N) + B (1)
In the above expression, S and N represent the settlement of ballast in
mm and the number of load cycles, respectively. Variables A and B are
Fig. 7. Final settlement of all ballast samples under cyclic load. the experimental factors presented in Fig. 8a to 8c. The experimental
variables A and B are functions of dry condition, geogrid characteristics
percentage in the samples, the efficiency and effectiveness of geogrid (including tensile strength, aperture size and location of geogrid),
decrease. In this condition, settlements of the reinforced samples and the ballast properties (i.e., rock type, compactness and grain size distribu­
unreinforced sample have not significant differences. The reduction of tion), and sand contamination. There are sudden increases and de­
geogrid efficiency originates from disturbance made by sand. Sand creases in values of the experimental variables A and B against sand
causes reduction of interlocking of geogrid with ballast particle. In fact, contamination. The sudden change is due to influence of sand contam­
small sand grains act as a lubricant and make ballast stones slide on ination on ballast behavior. That is, for a sample reinforced with certain
geogrid aperture and in turn, geogrid cannot trap and confine ballast geogrid and specified position of geogrid, changes of variables A and B
stones effectively. are because of different sand contents. Because sand small grains act
similar to lubricant, they have noticeable influence on settlement of
samples. Also, for samples reinforced with different geogrid positions, a
considerable changes in variables A and B is obtained because of

6
J. Sadeghi et al. Construction and Building Materials 362 (2023) 129654

Fig. 8. Values of A and B for ballast samples reinforced with geogrid.

changes in confinement effect induced by geogrid. The variation trend In Equation (2), ΔELoop and ATri refer to the energy damped in a
was obtained because position of geogrid has significant effects on loading–unloading cycle and the area of the right triangle marked in
ballast settlement. Fig. 10a, respectively. The energy dissipated in the cycle (ΔELoop) is
To assess the reliability and accuracy of the model, settlement results obtained by computing the area enclosed in a hysteresis loop.
of samples reinforced with G46 geogrid installed at 10 cm above ballast Also, damping ratio for ballast specimens can be obtained using
base (i.e., G46-H10 samples) were compared with results obtained from Equation (3) as under.
the prediction model. They are depicted in Fig. 9. According to Fig. 9, it
ΔELoop
is clear that the model can estimate ballast settlement precisely, having ζ= (3)
2πks2
less than 8 percent errors at the most.
In this Equation, k and s respectively denote the stiffness and the
3.3. Damping ratio and stiffness amplitude of settlement during a cycle [30]. In this study; damping ratio
is obtained from Equation (2), then stiffness is determined using Equa­
In accordance with [2], damping ratio of a soil sample subjected to tion (3) [47]. Hysteresis loops for specimens S14-WOGeo; S14-G34-H10
cyclic loading can be derived, using Equation (2). and S14-G46-H10 are presented in Fig. 10b to 10d. Values of stiffness
and damping ratio for the ballast samples at the end of cyclic load are
ΔELoop
ζ= (2) indicated in Fig. 11a and 11b, respectively. As illustrated in Fig. 11a, for
4πATri both reinforced and unreinforced samples when sand dosages become
higher than 7%, there is a reducing trend for stiffness results. As the
degree of sand contamination is small (i.e., 7%), the ballast matrix gets
more compacted. As a result, the ballast matrix becomes stiffer. On the
other hand, when the amount of sand exceeds 7%, the small sand grains
fill the gaps between coarse ballast aggregates and consequently, the
sand materials cause reduction of inter-particle friction of ballast.
For a specified amount of sand, reinforced samples demonstrate
more stiffness when compared with the unreinforced samples. This issue
takes place because geogrid restrains and confines ballast grains.
Therefore, ballast samples have less movement and sliding which lead to
an increase in the stiffness. In other words, due to interlocking/friction/
contact between ballast stones and geogrid aperture, ballast movement
is confined and ballast behaves in a stiffer manner. The results obtained
indicate that G24-H10, G34-H10 and G46-H10 specimens have 19%,
50% and 33% higher stiffness when compared with the unreinforced
specimen (i.e., WOGeo specimen) in clean condition. Also, in high
contamination cases (i.e., 36% sand content), G24-H10, G34-H10 and
Fig. 9. Contrast between settlements derived from the test (dashed line) and
G46-H10 specimens show 14%, 43% and 26% stiffer when compared
settlements predicted form the model (dotted line) for G46-H10 samples.

7
J. Sadeghi et al. Construction and Building Materials 362 (2023) 129654

Fig. 10. Hysteresis loops for samples S14-WOGeo, S14-G34-H10 and S14-G46-H10.

Fig. 11. Damping ratio and stiffness against sand contamination percentage in the final load cycle.

with unreinforced specimen. It indicates that with augmentation of sand Consequently, more amount of energy is dissipated. Actually, by pene­
content, effectiveness of geogrid reinforcement reduces gradually. tration of small sand particles into spaces among ballast stones, sliding
Results of damping ratio for ballast samples are presented in and rearrangement of particles are intensified because of reduction of
Fig. 11b. Because of viscoelastic characteristics of ballast material, me­ contact points and interlocking between ballast stones. By decrease in
chanical strain energy is stored and dissipated during cyclic loading. interlocking of ballast stones, sliding/rearrangement/rotation/move­
Dissipation of energy within ballast aggregates is obtained due to plastic ment of the aggregates become more which makes increases in damping
sliding and stone breakage [39]. As illustrated in Fig. 11b; damping ratio ratio. Fig. 11b shows that in the clean condition, damping ratio of
increases gradually for all ballast samples when sand percentage in­ WOGeo, G24-H10, G34-H10 and G46-H10 specimens are 0.20, 0.18,
creases. As ballast gets contaminated with sand, plastic sliding of ballast 0.14 and 0.16, respectively. For these specimens, in highly contami­
grains intensifies (due to lubrication effect of small sand grains). nated conditions, damping ratios are increased, reaching to 0.25, 0.23,

8
J. Sadeghi et al. Construction and Building Materials 362 (2023) 129654

0.19 and 0.21, respectively. As illustrated in Fig. 11b, damping ratio of cushion substance (i.e., soft layer) causing reduction of interaction,
ballast samples are decreased due to geogrid reinforcement. This friction and interlocking between ballast stones. As a result of decrease
reduction occurs because geogrid traps ballast grains into its openings. in contact between ballast particles, abrasion and fracture of particles
Consequently, displacement, rearrangement and sliding of the grains are are decreased considerably. According to Fig. 12b, when sand contam­
limited, causing reduction in dissipation of energy in the reinforced ination is 0%, BBI of G24-H10, G34-H10 and G46-H10 specimens are
samples. 7%, 23% and 12% less than unreinforced sample, respectively. Results
shown in Fig. 12b indicate that regardless of sand content, there is a
reduction in the BBI because of geogrid reinforcement. According to the
3.4. Ballast particle degradation literature, as ballast gets reinforced with geogrid, the intensity of inter-
particle contact stresses are decreased [25,26] and consequently the
Ballast particles are deteriorated under cyclic loads due to the fracture and attrition of particles become less when compared to unre­
changing mechanisms in ballast properties including (i) breakage and inforced ballast [27]. In other words; as reposition and displacement of
splitting of aggregates, (ii) attrition of sharp corners and asperities, (iii) ballast particles are restricted by the inclusion of geogrid within the
grinding of angularities and roughness, and (iv) turning particles to ballast layer, the possibility of the particle crushing becomes less. This
relatively rounded and semi-smooth grains. These four mechanisms restriction leads to a drop in breakage of aggregates. However as sand
make ballast stones have less inter-particle friction and interlocking. percentage reaches 36%, BBI of G24-H10, G34-H10 and G46-H10
This causes reductions in strength and stiffness of ballast matrix. Due to specimens are 4%, 12% and 8% lower than that of the unreinforced
presence of small sand grains between ballast stones, the decrease in sample, respectively. It indicates that by augmentation of sand content
contact/friction/interlocking within the stones is intensified remark­ in ballast samples, effectiveness of geogrid reduces.
ably. [27,40,50,22]. To study the effect of sand contamination on
degradation (i.e.; particle breakage) of ballast samples, ballast breakage 4. Effective ballast reinforcement using geogrid
index (BBI) is used. To obtain BBI of ballast samples, Equation (4) is used
[21]. Analysis of the results reveals that efficiency of geogrid is dependent
A on the aperture size and position of geogrid in the ballast. The results
BBI = (4) obtained here were used to derive the effective/suitable aperture size
A+B
and position of geogrid in contaminated ballasts.
In this Equation, A is the area between the grain size distribution
Comparison of the results obtained for settlement, stiffness and
curves before test and after cyclic test. B is the area of the zone between
ballast degradation (as illustrated in Figs. 6, 11 and 12) indicates that
the arbitrary curve of greatest breakage and the ultimate grain size
application of a 34 × 34 geogrid leads to less settlement, less reduction
distribution curve as indicated in Fig. 12a. To obtain BBI of samples, at
of BBI and more stiffness when compared to 24 × 24 and 46 × 46
first, sand materials were separated from ballast using #18 sieve (i.e., 1
geogrids, regardless of geogrid positions. This is due to d50 of ballast
mm sieve) because as depicted in Fig. 2, sand materials pass through this
material used in this study was 32 mm. Because ballast material has d50
sieve. For each sample the weight of sand added to ballast is certain. In
of 32 mm, a great number of ballast stones are trapped in 34 × 34 grids.
turn, if sand weight is increased, the augmentation is caused due to
Note that d50 refers to the size of particles in which 50% of ballast ag­
ballast breakage and crushing. For all the tests, the weight of grains
gregates are smaller. Using 34 × 34 geogrid, the interaction of ballast
passing through #18 sieve (1 mm) is almost equal to the weight of sand
particles and geogrid apertures is more when compared to the two other
added to the ballast sample after the cyclic tests. It indicates that
geogrids. This issue makes ballast-geogrid composite have better me­
negligible amount of ballast aggregates have worn, been grinded or
chanical characteristics (i.e., less settlement and less BBI). For 24 × 24
broken into sizes less than 1 mm (i.e., dMax of sand materials).
geogrid, the opening is not large enough to interact well with the ag­
Results presented in Fig. 12b show that for all the ballast samples (i.
gregates which have the diameter more than 32 mm. As a result, the
e., for both the reinforced and unreinforced samples) there is a reducing
ballast samples reinforced with 24 × 24 geogrid have greater BBI and
trend for breakage index as sand dosage increases. The decreasing trend
settlement compared to those of samples with 34 × 34 geogrid. For 46 ×
is obtained because the number of contact points and interlocking be­
46 geogrid, the aperture is too large to interlock with a large amount of
tween ballast aggregates are decreased by infiltration of sand materials
ballast stones. In the other words, significant numbers of ballast
into the spaces between the aggregates. In fact, small sand grains act as a

Fig. 12. Results of BBI.

9
J. Sadeghi et al. Construction and Building Materials 362 (2023) 129654

aggregates are small enough to pass through the grids. Therefore, the expression was developed to estimate ballast settlement when ballast is
efficacy of 46 × 46 geogrid is not as much as 34 × 34 geogrid. There is reinforced with geogrid and contaminated with sand.
some interlocking between ballast particles and grids for the 24 × 24 It was found that to take the most advantages of ballast reinforcing
and 46 × 46 geogrids. However ballast degradation and settlement with geogrid, size of geogrid aperture should be close to d50 of ballast
(when reinforced with these geogrids) are less than those of unrein­ grains. Based on the results obtained, use of geogrid with 34 × 34 grids
forced samples. positioned at a height of 10 cm from the base of the ballast layer is a
The results obtained for settlement, stiffness and BBI (presented in practical and effective method of ballast strengthening against sand
Figs. 6, 12 and 13) reveal that when geogrid is positioned at a height of contamination under cyclic loading. This position does not cause any
10 cm from the bottom of the ballast layer, better behavior of ballast is problem for ballast maintenance operations (such as tamping).
derived. That is, more stiffness, less settlement and less BBI are obtained
compared to the results obtained for the samples with geogrid posi­ Declaration of Competing Interest
tioned at 20 cm above the bottom. As geogrid is positioned at 10 cm
from the bottom, there is a larger amount of ballast particles above the The authors declare that they have no known competing financial
geogrid, and therefore, there are more interaction and interlocking be­ interests or personal relationships that could have appeared to influence
tween apertures and particles. Furthermore, this location of geogrid the work reported in this paper.
makes no problem for regular maintenance operations of ballast layer (e.
g., tamping and stabilizing). Data availability
Considering the improvement of ballast behavior made by various
geogrids and maintenance required conditions, 34 × 34 geogrid posi­ Data will be made available on request.
tioned at a height of 10 cm from the bottom of ballast layer is the op­
timum method of reinforcing ballast in sandy areas. This confirms what References
has been reported by these authors based on direct shear tests and plate
load tests on contaminated ballasts [46]. It is worthy to note that to get [1] AS1085.14, Standards Australia: Railway track material. Part 14: Prestressed
concrete sleepers. Australian Standard : AS1085.14 - 2003, Sydney, Australia.
the highest benefit form reinforcement; the geogrid opening size should [2] ASTM D3999/D3999M, Standard Test Methods for the Determination of the
be in harmony with d50 of ballast particles. This treatment minimizes the Modulus and Damping Properties of Soils Using the Cyclic Triaxial Apparatus,
detrimental effects produced by sand invasion on ballast performance. ASTM International, West Conshohocken, PA, 2003, www.astm.org.
[3] X. Bian, W. Li, Y. Qian, E. Tutumluer, Analysing the effect of principal stress
rotation on railway track settlement by discrete element method, Géotechnique 70
5. Conclusions (9) (2020) 803–821.
[4] S.F. Brown, J. Kwan, N.H. Thom, Identifying the key parameters that influence
geogrid reinforcement of railway ballast, Geotext. Geomembranes 25 (6) (2007)
Sand contamination of railway lines deteriorates ballast mechanical 326–335.
properties such that ballast cannot perform its role in the railway track [5] L. Bruno, M. Horvat, L. Raffaele, Windblown sand along railway infrastructures: A
system. There is a lack of study into the improvement of ballast behavior review of challenges and mitigation measures, J. Wind Eng. Indust. Aerodyn. 177
(2018) 340–365.
under cyclic loading in sand contamination condition. This is addressed [6] S. Chawla, J.T. Shahu, R.K. Gupta, Design methodology for reinforced railway
in this research. To this end, an extensive experimental research was tracks based on threshold stress approach, Geosynth. Int. 26 (2) (2019) 111–120.
carried out, using a cyclic testing box machine. A large number of cyclic [7] Choi, Y.T., Lee, S.J., Hwang, S.H. and Jang, S.Y., 2016. Development of a ballast
settlement model using laboratory ballast box tests and field measurements. In
tests were made on clean ballast samples and samples contaminated
Proceeding of Civil Engineering Conference in the ASIAN Region CECAR7,
with different dosages of sands. The results obtained from tests on sand Hononolu, Hawaii.
contaminated ballast indicate that: [8] Code 301, 2005. Iranian national code 301: General technical specification of
superstructure of ballasted railway track. Vice-Presidency for Strategic Planning
and Supervision, Islamic Republic of Iran, Tehran, Iran, (in Persian).
• Sand contamination causes up to 85% increase in the ballast settle­ [9] P.A. Costa, R. Calçada, A.S. Cardoso, Track–ground vibrations induced by railway
ment, and traffic: In-situ measurements and validation of a 2.5 D FEM-BEM model, Soil Dyn.
• Stiffness of ballast reduces up to 63% when contaminated with sand. Earthquake Eng. 32 (1) (2012) 111–128.
[10] B.M. Das, Use of geogrid in the construction of railroads, Innovative Infra.
Solutions 1 (1) (2016) 1–15.
In order to improve the mechanical behavior of sand contaminated [11] B.M. Das, J. Penman, P.R. Anderson, Use of Geogrid In Subgrade-Ballast System Of
ballast under cyclic loads, geogrid reinforcement technique was used in Railroads Subjected To Cyclic Loading For Reducing Maintenance, California State
University, Sacramento, USA, 2010.
this research. In order to obtain an optimum method for geogrid rein­ [12] R.F. du Plooy, P.J. Gräbe, Characterisation of rigid polyurethane foam-reinforced
forcement of contaminated ballast, geogrid with various opening di­ ballast through cyclic loading box tests, J. South Afr. Institut. Civil Eng. 59 (2)
mensions and positions within the ballast sample were investigated. To (2017) 2–10.
[13] M. Esmaeili, P. Aela, A. Hosseini, Effect of moisture on performance of mixture of
achieve this goal, a parametric experimental investigation was carried sand-fouled ballast and tire-derived aggregates under cyclic loading, J. Mater. Civil
out. The aperture sizes of geogrids used in the tests were 24 mm, 34 mm Eng. 31 (2) (2019) 04018377.
and 46 mm in two series of the tests. These geogrid openings are selected [14] M. Esmaeili, J.A. Zakeri, M. Babaei, Laboratory and field investigation of the effect
of geogrid-reinforced ballast on railway track lateral resistance, Geotext.
based on d50 of ballast grains. For the first series of the tests, the geogrid
Geomembr. 45 (2) (2017) 23–33.
was installed at a height of 10 cm from the bottom of the ballast samples; [15] M. Esmaeili, P. Aela, A. Hosseini, Experimental assessment of cyclic behavior of
and for the second series, the geogrid was installed 20 cm above the sand-fouled ballast mixed with tire derived aggregates, Soil Dyn. Earthquake Eng.
ballast base. The following findings were obtained. 98 (2017) 1–11.
[16] M. Esmaeili, S. Farsi, A. Shamohammadi, Effect of rock strength on the degradation
of ballast equipped with under sleeper pad, Construct. Build. Mater. 321 (2022),
• The most effective geogrid reinforcement is the use of 34 × 34 126413.
geogrid, installed at the position of 10 cm above the bottom of ballast [17] M. Esmaeili, P. Namaei, Effect of mother rock strength on rubber-coated ballast
(RCB) deterioration, Construct. Build. Mater. 316 (2022), 126106.
layer. [18] C. Gallage, B. Dareeju, M. Dhanasekar, T. Ishikawa, Effects of principal stress axis
• The reinforcement increases ballast stiffness up to 50%, reduces rotation on unsaturated rail track foundation deterioration, Proc. Eng. 143 (2016)
settlement up to 75% and decreases ballast breakage up to 23%. 252–259.
[19] Gao L., Yin H., Xu Y., Shi S., Cai H., Wang X. 2022. The Settlement Characteristics
of Ballast Bed Based on Variable Boundary Ballast Box. In: Tutumluer E., Nazarian
Results obtained from cyclic tests have good agreement with those S., Al-Qadi I., Qamhia I.I. (eds) Advances in Transportation Geotechnics IV. Lecture
found from shear tests and plate load tests on ballasts (reported in the Notes in Civil Engineering, vol 165. Springer, Cham.
literature).
Using the experimental results obtained in this research, a novel

10
J. Sadeghi et al. Construction and Building Materials 362 (2023) 129654

[20] H. Huang, E. Tutumluer, W. Dombrow, Laboratory characterization of fouled [39] S.K. Navaratnarajah, B. Indraratna, Use of rubber mats to improve the deformation
railroad ballast behavior. Transportation Research Record, J. Transport. Res. Board and degradation behavior of rail ballast under cyclic loading, J. Geotech.
2117 (2009) 93–101. Geoenviron. Eng. 143 (6) (2017) 04017015.
[21] B. Indraratna, J. Lackenby, D. Christie, Effect of confining pressure on the [40] N.T. Ngo, B. Indraratna, C. Rujikiatkamjorn, Modelling geogrid-reinforced railway
degradation of ballast under cyclic loading, Géotechnique 55 (4) (2005) 325–328. ballast using the discrete element method, Transport. Geotech. 8 (2016) 86–102.
[22] B. Indraratna, H. Khabbaz, W. Salim, D. Christie, Geotechnical properties of ballast [41] N.T. Ngo, B. Indraratna, Improved performance of rail track substructure using
and the role of geosynthetics in rail track stabilization, Proc. Institut. Civil Eng.- synthetic inclusions: experimental and numerical investigations, Internat. J.
Ground Improv. 10 (3) (2006) 91–101. Geosynth. Ground Eng. 2 (3) (2016) 1–16.
[23] B. Indraratna, S.K.K. Hussaini, J.S. Vinod, On the shear behavior of ballast- [42] Qian, Y., Mishra, D., Tutumluer, E., Kwon, J., 2013. Comparative evaluation of
geosynthetic interfaces, Geotech. Testing J. 35 (2) (2011) 305–312. different aperture geogrids for ballast reinforcement through triaxial testing and
[24] B. Indraratna, N.T. Ngo, C. Rujikiatkamjorn, Behavior of geogrid-reinforced ballast discrete element modeling. In Proceedings of Geosynthetics, California, USA.
under various levels of fouling, Geotext. Geomembr. 29 (3) (2011) 313–322. [43] Rahman, A.J., 2013. Permeability, Resistivity and Strength of Fouled Railroad
[25] B. Indraratna, N.T. Ngo, C. Rujikiatkamjorn, Deformation of coal fouled ballast Ballast. Master’s thesis, University of Kansas, Kansas, USA, 101 pages.
stabilized with geogrid under cyclic load, J. Geotech. Geoenviron. Eng. 139 (8) [44] J. Sadeghi, A.R. Tolou Kian, A. Shater Khabbazi, Improvement of mechanical
(2012) 1275–1289. properties of railway track concrete sleepers using steel fibers, J. Mater. Civil Eng.
[26] B. Indraratna, S.K.K. Hussaini, J.S. Vinod, On the shear behavior of ballast- 28 (11) (2016) 04016131.
geosynthetic interfaces, Geotech. Testing J. 35 (2) (2012) 305–312. [45] J. Sadeghi, J.A. Zakeri, A.R. Tolou Kian, Effect of unsupported sleepers on rail track
[27] B. Indraratna, S.K.K. Hussaini, J.S. Vinod, The lateral displacement response of dynamic behaviour, Proc. Inst. Civil Eng.-Transport 171 (5) (2018) 286–298.
geogrid-reinforced ballast under cyclic loading, Geotext. Geomembr. 39 (2013) [46] J. Sadeghi, A.R. Tolou Kian, H. Ghiasinejad, M. Fallah Moqaddam, S. Motevalli,
20–29. Effectiveness of geogrid reinforcement in improvement of mechanical behavior of
[28] B. Indraratna, T. Ngo, C. Rujikiatkamjorn, Performance of ballast influenced by sand contaminated ballast, Geotext. Geomembr. 48 (6) (2020) 768–779.
deformation and degradation: laboratory testing and numerical modeling, Internat. [47] J. Sadeghi, A.R. Tolou Kian, M. Chopani, A. Khanmoradi, Effects of particle
J. Geomech. 20 (1) (2020) 04019138. gradations on cyclic behavior of ballast contaminated with sand, Construct. Build.
[29] Jayalath, G., Prasad, C., Gallage, C., Dhanasekar, M., Dareeju, B.S., Ramanujam, J., Mater. 342 (2022), 127943.
Lee, J., 2018. Pavement model tests to investigate the effects of geogrid as subgrage [48] Selig, E.T., Waters, J.M., 1994. Track Geotechnology and Substructure
reinforcement. In 12th Australian and New Zealand Young Geotechnical Management. Thomas Telford, London, UK, 446 pages.
Professionals Conference, Liu, H, Mills, P, Ruxton, N, & Mazengarb, C (Eds.), 6-9 [49] M. Sol-Sanchez, N.H. Thom, F. Moreno-Navarro, M.C. Rubio-Gámez, G.D. Airey,
November 2018, Hobart, Tasmania, Australian Geomechanics Society, Australia, 1- A study into the use of crumb rubber in railway ballast, Construct. Build. Mater. 75
8. (2015) 19–24.
[30] L. Jacobsen, Steady forced vibration as influenced by damping, J. Trans. ASME- [50] Y. Sun, C. Chen, S. Nimbalkar, Identification of ballast grading for rail track,
APM 52 (15) (1930) 169–181. J. Rock Mech. Geotechn. Eng. 9 (5) (2017) 945–954.
[31] A.S. Lees, J. Clausen, Strength envelope of granular soil stabilized by multi-axial [51] A.R. Tolou Kian, J. Sadeghi, J.A. Zakeri, Large-scale direct shear tests on sand-
geogrid in large triaxial tests, Can. Geotech. J. 999 (2019) 1–5. contaminated ballast, Proc. Inst. Civil Eng.-Geotech. Eng. 171 (5) (2018) 451–461.
[32] W.L. Lim, Mechanics of Railway Ballast Behaviour, University of Nottingham [52] A.R. Tolou Kian, J.A. Zakeri, J. Sadeghi, Experimental investigation of effects of
Nottingham, UK, 2004. sand contamination on strain modulus of railway ballast, Geomech. Eng. 14 (6)
[33] S. Liu, H. Huang, T. Qiu, J. Kwon, Effect of geogrid on railroad ballast particle (2018) 563–570.
movement, Transport. Geotech. 9 (2016) 110–122. [53] A.R. Tolou Kian, J. Sadeghi, J.A. Zakeri, Influences of railway ballast sand
[34] R.S. Malisetty, B. Indraratna, J. Vinod, Behaviour of ballast under principal stress contamination on loading pattern of pre-stressed concrete sleeper, Constr. Build.
rotation: Multi-laminate approach for moving loads, Computers and Geotechnics Mater. 233 (2020), 117324.
125 (2020), 103655. [54] Tutumluer, E., Huang, H., Hashash, Y.M.A. and Ghaboussi, J., 2007, September.
[35] McDowell GR, Lim WL, Collop AC, Armitage R, Thom NH, 2005. Laboratory Discrete element modeling of railroad ballast settlement. In AREMA Annual
simulation of train loading and tamping on ballast. InProceedings of the institution Conference.
of civil engineers-transport 2005 May (Vol. 158, No. 2, pp. 89-95). Thomas Telford [55] UIC Leaflets 713 R, International Union of Railways, 2004.
Ltd. [56] Yu, Z., Woodward, P.K., Laghrouche, O., Connolly, D.P., 2019. True triaxial testing
[36] G.R. McDowell, P. Stickley, Performance of geogrid-reinforced ballast, Ground of geogrid for high speed railways. Transport. Geotech. 100247.
Engineering 39 (1) (2006) 26–30. [57] X. Zhang, D. Thompson, H. Jeong, M. Toward, D. Herron, C. Jones, N. Vincent,
[37] S. Mehran Khoshoei, H. Mortazavi Bak, S. Mahdi Abtahi, S. Mahdi Hejazi, Measurements of the high frequency dynamic stiffness of railway ballast and
B. Shahbodagh, Experimental Investigation of the Cyclic Behavior of Steel-Slag subgrade, J. Sound Vib. 468 (2020), 115081.
Ballast Mixed with Tire-Derived Aggregate, J. Mater. Civil Eng. 33 (2) (2021) [58] G.S.F. Farag. Lateral spreading in basal reinforced embankments supported by pile-
04020468. like elements, Kassel University Press GmbH, 2008.
[38] NAUE GmbH & Co. K.G., 2019. www.naue.com.

11

You might also like