Improvement of railway ballast maintenance approach, incorporating ballast geometry and fouling conditions

Prévia do material em texto

Accepted Manuscript
Improvement of railway ballast maintenance approach,
incorporating ballast geometry and fouling conditions
J. Sadeghi, M.E.M. Najar, M. Mollazadeh, B. Yousefi, J.A. Zakeri
PII: S0926-9851(17)30703-6
DOI: doi:10.1016/j.jappgeo.2018.02.020
Reference: APPGEO 3451
To appear in:
Received date: 25 July 2017
Revised date: 18 February 2018
Accepted date: 20 February 2018
Please cite this article as: J. Sadeghi, M.E.M. Najar, M. Mollazadeh, B. Yousefi, J.A.
Zakeri , Improvement of railway ballast maintenance approach, incorporating ballast
geometry and fouling conditions. The address for the corresponding author was captured
as affiliation for all authors. Please check if appropriate. Appgeo(2018), doi:10.1016/
j.jappgeo.2018.02.020
This is a PDF file of an unedited manuscript that has been accepted for publication. As
a service to our customers we are providing this early version of the manuscript. The
manuscript will undergo copyediting, typesetting, and review of the resulting proof before
it is published in its final form. Please note that during the production process errors may
be discovered which could affect the content, and all legal disclaimers that apply to the
journal pertain.
https://doi.org/10.1016/j.jappgeo.2018.02.020
https://doi.org/10.1016/j.jappgeo.2018.02.020
https://doi.org/10.1016/j.jappgeo.2018.02.020
AC
C
EP
TE
D
 M
AN
U
SC
R
IP
T
 
1 
 
Improvement of railway ballast maintenance approach, 
incorporating ballast geometry and fouling conditions 
J. Sadeghi1*, M. E. M. Najar1, M. Mollazadeh1, B.Yousefi2, J. A. Zakeri1 
1 School of Railway Engineering, Iran University of Science and Technology, Narmak, 16844, Tehran, Iran 
2 Geophysics Department, Geotechnical & Strength of Material Study Center of Municipality, Tehran, Iran 
*Corresponding Author, Javad_sadeghi@iust.ac.ir 
Abstract: Ballast plays an important role in the stability of railway track systems. The 
effectiveness of the ballast in maintaining the track stability is very much dependent on its 
mechanical conditions. The available ballast maintenance approaches are mainly based on only 
track geometry conditions (such as track profile) which do not sufficiently reflect the ballast 
mechanical behaviors. That is, the ballast potential of degradation (i.e., ballast long term 
behaviors) has been omitted. This makes the effectiveness of the current ballast maintenance 
approach questionable, indicating a need for a more comprehensive and effective ballast 
conditions assessment technique. In response to this need, two ballast condition indices based 
on ballast geometry degradation (BGI) and the level of ballast fouling (BFI) as the main 
indicators of ballast mechanical behavior were developed. The BGI is a function of the standard 
deviations of track alignment, unevenness and twist. The BFI was developed based on the data 
obtained from the ground penetration radar (GPR). Making use of the new indices, a more 
reliable maintenance algorithm was developed. Through illustrations of the applicability of the 
new maintenance algorithm in a railway line, it was shown that the new algorithm causes a 
considerable improvement in the maintenance effectiveness and an increase in the life cycle of 
railway tracks by making more effective allocation of resources and more accurate maintenance 
planning. 
Keywords: Railway, Assessment technique, Quality index, GPR, Railway Ballast, 
Maintenance planning 
 
1. Introduction 
A conventional double-track line contains 3000 to 5000 m3 of ballast per kilometer, depending 
on the type of the track and the spacing of the lines. The economical handling and maintenance 
management of these huge quantities of material is one of the main concerns of railway 
industries. Ballast transfers train loads to the sub-ballast layer and plays significant roles in the 
lateral and longitudinal stability of railway track systems. The ballast has to have minimum 
required mechanical and geometry conditions in order to perform its role. As the track strength 
and stability are greatly dependent on appropriate functioning of the ballast (Misar, 2002), the 
ballast is considered as the main component in any railway track maintenance. The 
effectiveness of a ballast maintenance approach is dependent on the accuracy of the recording 
and assessment of ballast conditions (Anderson, Cunningham, & Barry, 2002; Caetano & 
Teixeira, 2015; Navikas, Bulevičius, & Sivilevičius, 2016; Nederlof & Dings, 2010). 
ACCEPTED MANUSCRIPT
AC
C
EP
TE
D
 M
AN
U
SC
R
IP
T
 
2 
 
In the current practice, the assessment of the ballast is made based on only track geometry 
conditions. The considered track geometry conditions include track profile, gauge, alignment, 
and twist. Various geometry indices have been developed based on these parameters (Javad 
Sadeghi & Askarinejad, 2010). The track geometry parameters are usually recorded by track 
recording cars which run with a maximum speed of 120 km/h and even more (Auer, 2013; 
Guler, 2014; Z. Li, Lei, & Gao, 2016; J Sadeghi, 2010; Vale & Ribeiro, 2014; Van der 
Westhuizen, 2012). The currently used indices do not reflect the cause of the geometry defects. 
That is, the track geometry parameters refer to track serviceability generally related to the track 
conditions required to safely convey passing traffic instantly. They are often a poor indicator 
of future performance due to changing conditions and nonlinear stress–strain behavior of the 
track substructure (D. Li, Hyslip, Sussmann, & Chrismer, 2016). Although several attempts 
have been made in order to take into account the ballast structural conditions (JM Sadeghi & 
Askarinejad, 2011; Uzarski, Darter, & Thompson, 1993), they have failed since their 
procedures are time consuming and they are limited to only surface visible defects. Moreover, 
manual pitting tests or analyses of logs (specimens) have not been considered as they are too 
costly and time consuming. The conventional ballast assessment methods are based on visual 
inspections of tracks on-site in which samples of ballast materials are taken and analyzed in a 
lab for particle size distribution. Addressing the limitation of the current methods, this paper 
presents a non-contact approach, capable of performing more accurate and effective inspection 
of ballast conditions. 
A review of the literature indicates that there is an urgent need for a more economical, efficient 
and non-destructive method (Frangopol & Liu, 2007; Kim, Ahn, & Yeo, 2016; Orlando, 
Cardarelli, Cercato, De Donno, & Di Giambattista, 2017) of continuous mechanical condition 
monitoring of ballast and a more reliable ballast conditions index by which the ballast short 
and long term behaviors can be quantified (Camargo, Edwards, & Barkan, 2011; De Bold, 
O’Connor, Morrissey, & Forde, 2015; Jiménez-Redondo, Escriba, Benítez, Cores, & Cáceres, 
2014). In response to this need, in this research, new indices for ballast condition evaluation 
were established based on the ballast level of contamination (which reflects the ballast 
mechanical behavior) and the ballast layer geometry (which is the bases of ballast stability 
conditions). 
In this paper, the structural index was established based on statistical analysis of the data 
obtained from automated GPR. The geometry index was developed based on the track 
geometry parameters including profile, alignment and twist obtained from a track recording 
car. Making use of the new indices, an improved ballast maintenance algorithm was developed. 
Applicability and effectiveness of the new algorithm in the maintenance activities (tamping 
and cleaning) were illustrated by applying the new approach in a railway line. 
2. Development of ballast fouling index 
According to the literature (Anbazhagan, Dixit, & Bharatha, 2016; Fontul, Fortunato, De 
Chiara, Burrinha, & Baldeiras, 2016; Selig & Waters, 1994), the main parameter indicatingthe 
structural conditions of the ballast is the ballast degree of fouling (contamination of ballast with 
fine materials). Aggregate breakage, infiltration of fine particles from the underlying subgrade 
layer (i.e., pumping effect), and intrusion of fine materials from the ballast surface are the main 
sources of ballast contamination (Nimbalkar, Indraratna, Dash, & Christie, 2012; Tennakoon 
& Indraratna, 2014). In this research, the ballast fouling (contamination) was taken as the main 
indicator to develop a ballast structural condition index. This was made based on the data 
obtained from the Ground penetration radar (GPR). The GPR provides a rapid, nondestructive 
ACCEPTED MANUSCRIPT
AC
C
EP
TE
D
 M
AN
U
SC
R
IP
T
 
3 
 
measurement of ballast conditions (Anbazhagan, Lijun, Buddhima, & Cholachat, 2011; Clark, 
Gordon, Giannopoulos, & Forde, 2004; Hugenschmidt, 2000; Hyslip, Smith, Olhoeft, & Selig, 
2003; Khakiev, Shapovalov, Kruglikov, & Yavna, 2014; Olhoeft & Selig, 2002; Pilecki et al., 
2017; Sharpe, 2000). In this technique, the ballast contamination level as well as the ballast 
depth can be obtained. In the GPR technique the absorption of the GPR waves/signals 
(transmitted to the ballast) increases as the contamination of the ballast increases. This is the 
main principle in the GPR technique to determine the level of ballast contamination 
(Manacorda & Simi, 2012). Despite extensive studies on the ballast fouling benchmark based 
on GPR technology (Brough, Stirling, Ghataora, & Madelin, 2003; De Bold et al., 2015; 
Gallagher, 1999), there is still a lack of numerical rating methods and applicable indices for 
the evaluation of ballast mechanical conditions for maintenance purposes. 
In order to develop a ballast fouling index (BFI), four steps were taken. First, GPR 
measurements were performed in a laboratory to make correlation between the levels of fouling 
(contamination degree) and the GPR output. Second, field boring tests were carried out to drive 
the GPR wave velocity for various types of ballast. Third, the method of deriving the level of 
ballast fouling was developed based on an image processing of the GPR data. Finally, the 
ballast fouling index was developed in a form of a mathematical expression by which the 
intensity level of the ballast fouling can be derived. 
2.1. Ballast fouling level based on GPR 
There is no definition available for ballast fouling levels based on the GPR outputs. In order to 
use the GPR technique, there is a need to make correlation between the ballast level of fouling 
(as defined in the literature) and the GPR output. For this purpose, several samples of ballast 
with different amounts of fouling were made. Through laboratory tests, correlations were 
developed between GPR data and the amounts of fouling. The accuracy of this procedure 
depends on the fine material used in the process of making samples. The procedure had a good 
level of accuracy since the ballast were obtained from the field and the type and the amount of 
fine materials were chosen based on the amount of the ballast aggregate breakage in the field 
(the cause of ballast contamination). The fine materials (i.e., the fouling materials) were chosen 
based on the result of field ballast screening. The ballast samples were provided according to 
Iranian standard No. 301 (IMRT, 2005). They were set in a nonferrous chamber with 1m length 
and 1m width. The chamber had 4 plastic frames with a depth of 10 cm (Fig-1). 
The ballast samples were contaminated with various fouling degrees. The amount of fouling 
was made based on the Selig ballast fouling gradation (Equation (1) and Table 1) which has 
been widely used in the world (Selig & Waters, 1994). In this equation, P4 and P200 are the 
mass percentages of particles less than 4.75 mm and 0.075 mm in diameter, respectively. 
 
(1) 
 
Table 1: Selig fouling degree (Selig & Waters, 1994) 
Clean Moderately clean Moderately Fouled Fouled Highly Fouled 
<1 1≤ <10 10≤ <20 20≤ <40 ≥40 
The sieves Numbers 4 and 200 were used to screen the materials. After placing the ballast 
aggregates in 5 cm height, they were compacted by a steel rod to achieve the assigned level of 
compaction (as in the field). 
ACCEPTED MANUSCRIPT
AC
C
EP
TE
D
 M
AN
U
SC
R
IP
T
 
4 
 
 
 
 
(a) Separation of subgrade and the 
ballast by plastic sheet 
(b) Filling with ballast aggregates (c) Compaction of ballast 
Fig 1. Preparation of samples with various fouling degrees 
 
The antenna type was selected such that the 40 cm thickness ballast layer can be scanned with 
sufficient precision. The samples of ballast layer consisted of coarse aggregates and had 40 
centimeters depth. A 20 cm of sandy clay layer was made under the ballast layer to simulate 
the field subgrade. A plastic sheet was used to separate the subgrade from the ballast aggregate. 
A scheme of laboratory installation is presented in Figure (2). 
Table 2: Specimens condition for laboratory scanning 
Sample Fouling amount (bottom layers) Fouled Mass - kg 𝑃4 + 𝑃200 
1 Clean ballast 0 0% 
2 Moderate clean 20 10% 
3 Moderate fouled 40 20% 
4 Fouled 60 30% 
5 Highly fouled 80 40% 
 
 
Fig 2: Semantic view of laboratory tests 
Five samples with different fouling degrees were prepared in the laboratory. Their details are 
presented in Table (2). The top 20 cm of the samples was filled with clean ballast. When the 
ACCEPTED MANUSCRIPT
AC
C
EP
TE
D
 M
AN
U
SC
R
IP
T
 
5 
 
samples were made ready, the radar antenna was dragged along the surface of each ballast 
specimen. This is shown in Figure (3). 
 
Fig 3: Dragging GPR antenna along surface of specimen 
In this research, a 2-GHz antenna was used. The wave amplitude and the amounts of the 
reflection of the wave transmitted from the radar in the samples were obtained. Two laboratory 
samples of the wave reflection with different fouling degree are presented in Figure 4. 
 
 
(a) (b) 
Fig 4: Two instances of wave reflection obtained from laboratory test; top layer is 20 cm 
clean ballast; the bottom layer is 20 cm ballast aggregate which included fine materials 
(a) with 10% fouling degree (b) with 40% fouling degree 
 
The more fouling of the ballast, the less reflection or transmutation of the electromagnetic 
waves is obtained. Using the SPSS (Radan) software, different colour spectrums were produced 
based on the amplitude of wave responses, so that each colour represents the level of ballast 
fouling along the ballast depth and length. 
The data for each run was imported into the RADAN6.6 (a commercial software) and the 
contour plots of the radar data were produced. The procedures to produce color-coded data are 
indicated in the following flow-diagram (Figure 5). 
ACCEPTED MANUSCRIPT
AC
C
EP
TE
D
 M
AN
U
SC
R
IP
T
 
6 
 
1-Raw data
6- Horizontal 
moving average
3- Background 
removal
5- Hilbert 
transform
2- Time Zero 
correction
8- Color transform
7- Vertical low 
pass filter
4-Gain restoration
 
Fig 5: GPR data processing (Roberts, Al-Audi, Tutumluer, & Boyle, 2008) 
In the color spectrum section of the software, there is an option to define a color for any specific 
range of the signals received. In this study, the colors were defined from green to black. The 
green color was defined for the maximum range of the wave amplitude (clean ballast) and black 
was for the lowest range of wave amplitude (highly fouled ballast). A sample of color spectrum 
is presented in Fig. 6. In this figure, the horizontal axis indicates the length of the ballast profile 
along the track (in each fifty meters) and the vertical axis is the track depth in centimeter (from 
the top surface of the ballast toward the bottom of the underneath layers). 
 
 
Fig 6: Color spectrum produced by analysis of field GPR data 
 
Through spectrum analyses of the results, four colors were selected as representativesof the 
four fouling ranges (indicated in Table 1). Based on the fouling gradation suggested by Selig 
(Table (1)), a color spectrum corresponding to the ranges of contamination was assigned to 
represent the ballast fouling levels. They are presented in Table 3. 
Table3: Color spectrum definition for different ballast fouling levels 
Highly fouled 
(30-40) 
Fouled 
(20-30) 
Moderately 
fouled 
(10-20) 
Clean 
(0-10) 
Fouling 
Black Red Yellow Green 
Color 
spectrum 
 
 
 
 
ACCEPTED MANUSCRIPT
AC
C
EP
TE
D
 M
AN
U
SC
R
IP
T
 
7 
 
2.2. GPR wave velocity in contaminated ballasts 
To drive the ballast layer thickness from the GPR output, there is a need to obtain the GPR 
wave velocity (Santos & Teixeira, 2017) in the ballast materials. For this purpose, 12 points 
along the railway track (with various ballast contamination levels) were selected. These points 
were closed to the sleeper ends where the GPR antennas scanned. 
 
 
Fig 7: Pitting test in field track 
 
Table 4: GPR wave velocity in contaminated ballast 
Point 
Number 
Ballast thickness 
measured (cm) 
Contamination level 
(Selig Eq.1) measured 
Wave speed 
computed (Eq. 2) 
m/ns 
1 42 10.4 0.115 
2 40 13.3 0.110 
3 32 15.5 0.107 
4 32 11.7 0.111 
5 42 0.6 0.139 
6 30 1.2 0.134 
7 12 2.7 0.128 
8 37 8.5 0.122 
9 45 4.2 0.121 
10 50 3.6 0.124 
11 58 7.3 0.122 
12 47 4.8 0.125 
 Rounded Average Speed (m/ns) 0.1215 
 
The actual thicknesses of the ballast were obtained by making bores in the ballast (Figure (7)). 
For a particular point (x), having the wave traveling time (T) and the thickness of the ballast 
(S), the wave velocity can be derived by dividing (S) by half of (T). The results obtained are 
summarized in Table (4). The GPR wave speed was calculated by averaging the speeds 
recorded from the 12 points. The average speed of GPR wave in the ballast was 0.1215 m/ns. 
 
2.3. Numerical index of ballast mechanical condition 
As indicated above, the ballast mechanical quality can be assessed by the spectrum analysis of 
the data obtained from GPR. The color spectrum mapping of the ballast layer contamination 
cannot be directly used for the maintenance management purposes, unless it is converted into 
a numerical rating. Using the MATLAB software, the area of each color was presented in 
ACCEPTED MANUSCRIPT
AC
C
EP
TE
D
 M
AN
U
SC
R
IP
T
 
8 
 
percentage forms. For this purpose, thickness of the ballast layer of each segment was 
computed based on the method described in Section 2. The average depth in each segment was 
considered, and in turn the percentage of each colored area was obtained. This is indicated in 
Figure (8). 
 
 
Fig.8: Computation of GPR spectrum areas 
The GPR fouling index is defined based on the colored areas (i.e., fouling intensity) which 
represent the ballast fouling condition. It is presented in the following form: 
 
(2) 
where, BFI is the new ballast fouling index, G,Y, R, and B stand for the percentage of green, 
yellow, red and black colors, representing fouling degree levels of clean, moderate fouled, 
fouled, and highly fouled ballasts, respectively. The coefficients were derived from the mid-
band of each fouling degree based on Table (5). 
Table 5: Augment factor determination 
Color Green Yellow Red Black 
AF 
Fouling Degree 0-10 10-20 20-30 30-40 
Limit of 
States 
State 1 100 0 0 0 5 
State 2 0 100 0 0 15 
State 3 0 0 100 0 25 
State 4 0 0 0 100 35 
 
The ballast fouling index can be expressed by the following concise form: 
 
(3) 
Where BFI is the ballast fouling index, Ai is the area for the i
th color, representing the intensity 
of the fouling, AF is an augment factor for highlighting the fouling degree, and n is the number 
of colors (1 to 4). Classification of the ballast condition based on the new index in comparison 
with that of Selig (Selig 1994) is presented in Table (6). In addition to the ballast fouling, there 
might be mud holes or drainage issues in some parts of the track. These were considered and 
ACCEPTED MANUSCRIPT
AC
C
EP
TE
D
 M
AN
U
SC
R
IP
T
 
9 
 
simulated in the laboratory by adding moisture and fine materials to the samples. Mud holes or 
drainage issues are in the black color range due to the reduction of wave’s energy. When a 
black color becomes visible (indicting highly contamination of the ballast), the location of the 
mud holes or water trapped (in the ballast pocket) can be detected. Also variability of the 
subgrade and foundation properties along the track was detected by changes in the reflections 
of the GPR radiations. 
Table 6: Ballast fouling index classification 
Ballast Condition 
Fouling Number from 
Eq. (3) 
Selig Index 
Clean <10 <1 
Moderately clean 10-15 1<<10 
Moderately fouled 15-20 10<<20 
Fouled 20-25 20<<40 
Highly fouled >25 >40 
 
3. Ballast Geometry Index 
Various track geometry condition indices have been developed. They include roughness index, 
fractal analysis index, and space curve length index in the USA, W5-parameter in Austrian 
railway, TGI in Indian Railway, Q index in Sweden National Railway, and J index in 
Poland(Berawi, 2013; Berawi, Delgado, Calçada, & Vale, 2010; J Sadeghi, 2010). Among 
these indices, the J index is more practical and more easy to use for the maintenance purposes 
(Scanlan, Hendry, & Martin, 2016), This index evaluates the track condition with respect to 
the standard deviations of track geometry parameters including twist (T), alignment (A), gauge 
(G) and unevenness (U) (Madejski & Grabczyk, 2002). 
The ballast geometry index was adapted from the J index. Gauge parameter has no meaningful 
relation with the ballast layer geometry irregularities (Lichtberger, 2005). On the other hand, 
deviations of the profile, alignment and twist have dominant effects on the geometry condition 
of the ballast layer. Therefore, the new geometry index for the ballast was obtained from the 
elimination of the gauge parameter from the J index. Therefore, the new index has the 
following format; 
 
 
(4) 
where BGI is the ballast geometry index for a track segment with a certain length (200 meter 
according to EN13848-6 2014),SA represents the standard deviation of alignment, SU is the 
standard deviation of unevenness and STW is standard deviation of twist. 
The standard deviations of the profile, alignment and twist are obtained from a track recording 
machine. These three parameters indicate the projection of the track surface onto vertical and 
horizontal planes which specify ballast layer position in the space. The average of rails 
alignment represents center of the ballast layer in the horizontal plane. In the vertical plane, 
average of longitudinal rails unevenness can represent ballast layer elevation. Twist is used to 
combine transverse vertical plane to represent changing rate of ballast position along the track. 
The allowable limits of J index after which the ballast needs maintenance and repair actions 
for various train speeds are presented in Table (7). 
ACCEPTED MANUSCRIPT
AC
C
EP
TE
D
 M
AN
U
SC
R
IP
T
 
10 
 
 
 
Table 7: Allowable ballast geometry index according to the J coefficient (Berawi et al., 2010) 
160 140 120 110 100 90 80 40 30 
Speed 
(km/h) 
2.0 2.8 4.0 4.9 5.5 6.2 7 11.0 12.0 J limit 
 
4- New ballast maintenance algorithm 
Simultaneous consideration of ballast geometry defects and ballast structural conditions 
enables engineers to conduct an integrated assessment of ballast conditions. For this purpose, 
a new ballast maintenance algorithm was developed, taking into account the BGI and the BFI. 
A flow-diagram of the new algorithm is presented in Figure (9).Based on this new maintenance 
approach, the maintenance planning is set based on both mechanical and geometry conditions 
of the ballast In this approach, three levels of threshold were defined for ballast maintenance; 
(i) safelimit (SL) which indicates the regularly planned maintenance operations, (ii) tamping 
limit (TL) which requires tamping of ballast layer and (iii) ballast cleaning limit (BCL) which 
requires ballast screening and renewal. These threshold levels are used to determine the 
required maintenance actions and the critical track segments which need urgent repairs. 
Track ballast inspection
Structural surveyGeometry survey
Ground penetration 
radar (GPR vehicle)
Track recording car 
(EM 120)
Segmentation and 
Data processing
Computing ballast 
fouling index (BFI)
Computing ballast 
geometry index (BGI)
Track ballast maintenance planning 
Threshold controlling 
based on Table (6) and (7) 
Regular MaintainingTampingBallast cleaning
 
Fig. 9: New ballast maintenance algorithm 
ACCEPTED MANUSCRIPT
AC
C
EP
TE
D
 M
AN
U
SC
R
IP
T
 
11 
 
The BGI indicates the safe limit (SL) and the tamping limit (TL) while the BFI indicates the 
ballast cleaning limit (BCL). Urgent tamping activities should be made when BGI passes the 
allowable limits according to Table (7). Based on the ballast fouling index derived from GPR 
output, the requirement for ballast cleaning can be identified using Table (6).For instance, the 
immediate ballast cleaning activities should be performed when GPR fouling index is more 
than 25 (highly fouled degree) as indicated in Table (6). The new maintenance algorithm 
improves the current ballast maintenance approach by considering the level of ballast 
contaminations and the ballast layer geometry deviations simultaneously. 
5-Application of New Developed Ballast Maintenance Approach 
In order to illustrate the effectiveness of the new proposed approach, its application in a railway 
line was discussed and evaluated. 
5-1. Site Description 
17 km railway line of the Iranian railway network between Mo'men-Abad and Azna stations in 
the west of Iran was selected (Fig.10). This block is located in a main line used for passenger 
and freight trains. The annual passing load is about 3.5 million gross tones (MGT)with the 
maximum train speed of 110 km/h. The test site was a ballasted track with 1435 mm track 
gauge. It consists of UIC 60 rails and B70 concrete sleepers with a center to center sleeper 
spacing of 60 cm. The line slope in this block is mainly less than 0.7 %. 
 
Fig 10: Location of the test section in Arak district of Iran railway network 
The ballast in the field is made up of crushed granite. The region is subjected to harsh climate 
conditions; winter snowfall and extreme temperatures ranges from -30 °C in the winter to 40 
°C in the summer. The line was divided into segments of 200 m length irrespective of the track 
structure properties and curves locations. The results were obtained from data collection of 80 
segments along the block. 
 
5.2. Data Collection and Analysis 
According to the literature, the ballast conditions degradation happens mainly under the rails 
and at the ends sides of the sleepers (Nurmikolu, 2005). Therefore, a pair of 2D GPR antennas 
was used for each end of the sleepers (Fig 11a). In this research, a 2GHz air-coupled antenna 
(which was the same as that in the laboratory tests) was mounted on the rail vehicle and 
suspended 40 cm above the sleeper surface (Fig 11b). The data collection rate was controlled 
by a digital measurement instrument (DMI). Data collection operation was carried out with a 
ACCEPTED MANUSCRIPT
AC
C
EP
TE
D
 M
AN
U
SC
R
IP
T
 
12 
 
rail car with the speed of 60 km/h and 10 scans per meter. The data was collected when moving 
longitudinally along the track. The GPR antenna recorded the data in every 30 cm. The 
configurations of the GPR for manual dragging (in the lab) and the rail car were different such 
that each configuration suits the method of scanning. 
 
 
 
(a) (b) 
Fig 11: GPR rail-vehicle setup with two 2-GHz antennas 
In order to calculate the GPR fouling index for a segment (with 200 m length),the ballast zone 
was specified using two white lines from zero depth up to the bottom of the ballast layer. 
Through GPR data processing (image processing), as described in Section 2,the ballast fouling 
indexes for the segments were obtained. For instance, the color spectrum (and the computed 
BFI) of three different segments are presented in Figure (12). 
Black Red Yellow Green 
 
11.21 % 42.52 % 39.89% 6.47% 
 
 
Black Red Yellow Green 
 
3.38% 31.75% 63.6 % 1.81% 
 
 
Black Red Yellow Green 
 
0.9% 7.46% 58.8% 32.85% 
 
Fig. 12: Computation of fouling index for three different segments 
ACCEPTED MANUSCRIPT
AC
C
EP
TE
D
 M
AN
U
SC
R
IP
T
 
13 
 
Figure (13) presents the BFI along the track obtained for the Mo'men Abad-Azna block 
between kilometer benchmarks 402 and 418 km. 
 
Fig. 13: BFI obtained for 80 measured segments 
As illustrated in Figure (13) and Table (5), the 17-km block includes 6%of clean and 
moderately clean ballast, 28% of moderately fouled ballast, 54% of fouled ballast, and 12% of 
highly fouled ballast. 
The track geometry parameters were obtained, using a track geometry recording machine called 
EM 120. Based on the 25 cm interval of measurements, standard deviations of vertical (V) and 
horizontal (H) irregularities for all 200-m track segments were computed. The chord length in 
the measurement for the unevenness and alignment was 10 m. It was 5 m for the twist. The 
unevenness and the alignment were computed by averaging the amounts obtained for the left 
and right rails as indicated under. 
 
(5) 
 
The standard deviations of three main parameters in the BGI for all the segments were derived. 
They are presented in Figure (14a, b and c). 
 
 
(a) Alignment (b) Unevenness 
0
5
10
15
20
25
30
1 6 11 16 21 26 31 36 41 46 51 56 61 66 71 76 81
Fo
u
lin
g 
in
d
ex
Segment No.
GPR Fouling Index
0.0
2.0
4.0
6.0
8.0
10.0
1 6 11 16 21 26 31 36 41 46 51 56 61 66 71 76 81
SD
 (
m
m
)
Segment No.
Unevenness Standard Deviation
ACCEPTED MANUSCRIPT
AC
C
EP
TE
D
 M
AN
U
SC
R
IP
T
 
14 
 
 
(c) Twist (d) Thickness 
Fig. 14: Standard deviations of ballast layer geometry parameters 
The GPR waves traveling time were recorded at the boundaries of the ballast and subgrade in 
every 30 cm along the track and the ballast thickness at each 30 cm of the track was computed 
(based the method described in Section 2). The standard deviations of the data (Figure 14-d) 
and the average ballast thickness for each track segment were obtained. 
Based on Equation (5), the ballast geometry index (BGI) was derived. The results obtained for 
80 segments are presented in Figure (15). 
 
 
Fig 15: Ballast geometry index along the track (80 segments) 
 
Taking into account the maximum train speed of 110 km/h (as indicated by the railway 
authorities), the results indicates that 43% of the track length has exceeded the allowable BGI 
limit according to Table (7). It means that less than half of the block need ballast tamping. The 
average ballast thickness in the test zone was around 37 cm. It indicates that there is no need 
to bring new ballast materials to the site for the ballast tamping (i.e., the ballast can be 
rearranged along the block during the taming). 
The critical sections in which track need urgent maintenance actions are presented in Figure 
(16). Figure (16-a) indicate track sections in which BFI exceeds 25; and therefore, they need 
ballast cleaning. Figure (16-b) presents the sections in which BGI is more than 4.9 and 
therefore, they need tamping to gain acceptable ballast layer geometry profile. Fig (17) 
represents critical zones in which the ballast needs urgent cleaning and tamping. The results 
obtained indicate that the new proposed maintenance approach clearly shows the urgency of 
the maintenance action or distinguishes various types of repairactions required. 
0.0
2.0
4.0
6.0
8.0
10.0
1 6 11 16 21 26 31 36 41 46 51 56 61 66 71 76 81
SD
 (
m
m
)
Segment No.
Twist Standard Deviation
0.0
2.0
4.0
6.0
8.0
1 6 11 16 21 26 31 36 41 46 51 56 61 66 71 76 81
SD
 (
cm
)
Segment No.
Ballast Thickness Standard Deviation
0.0
1.0
2.0
3.0
4.0
5.0
6.0
7.0
8.0
1 6 11 16 21 26 31 36 41 46 51 56 61 66 71 76
B
G
I G
ra
d
e
Segment No.
Ballast Geometry Index (BGI)
ACCEPTED MANUSCRIPT
AC
C
EP
TE
D
 M
AN
U
SC
R
IP
T
 
15 
 
 
 
Fig. 16: Critical segment based on BFI (a) and BGI (b) 
 
 
Fig. 17: Sections in which ballast repaired maintenance along the track 
 
As illustrated above, the new proposed algorithm provides a more précised decision between 
tamping, cleaning or replacement of the ballast. It eliminates short interval tamping actions by 
distinguishing the ballast cleaning time. Moreover through the new algorithm, the critical zones 
for which urgent repair action are required, are flagged. This ensures track safety and less long 
term maintenance cost. The deterioration rate of ballast layer conditions can be derived by 
monitoring the track ballast conditions (computing the new indices) in specific period (between 
two intervals). 
6- Conclusions 
The conventional ballast maintenance approach (maintenance planning) is based on only track 
geometry conditions. That is, it does not indicate the ballast mechanical conditions (i.e., the 
potential of ballast degradation). In other words, the current practice concentrates on the ballast 
short term behaviour and its long term behaviour has been omitted. While the track geometry 
conditions may appear normal, the track can be on the verge of failing due to a sever ballast 
fouling. It means that track geometry by itself cannot lead to precise maintenance planning. 
Addressing this limitation, the current ballast maintenance approach was improved by 
developing a new maintenance algorithm which takes into account both short and long terms 
behaviour of the ballast. For this purpose, two ballast conditions indices called BFI and BGI 
are developed. A ballast fouling index (BFI) was developed to indicate the level of ballast 
contamination which has been known as the main cause of ballast mechanical deterioration. 
This index was developed by analysing the results obtained from GPR machines. For this 
purpose, compressive laboratory tests were carried out to derive correlation between ballast 
contamination levels and the output of GPR machines. The correlation was used to drive a 
mathematical expression for the ballast contamination (BFI). A ballast geometry index (BGI) 
was developed based on the ballast geometry deviations in the vertical and lateral directions. It 
was made by analyses of the track geometry parameters deviations obtained from a track 
recording car. The main futures of these indices are: (1) they directly reflect both short and 
1 3 5 7 9 11 13 15 17 19 21 23 25 27 29 31 33 35 37 39 41 43 45 47 49 51 53 55 57 59 61 63 65 67 69 71 73 75 77 79
Segment No.
GPRFI- Critical Zones
1 3 5 7 9 11 13 15 17 19 21 23 25 27 29 31 33 35 37 39 41 43 45 47 49 51 53 55 57 59 61 63 65 67 69 71 73 75 77 79
Segment No.
BGI-Critical Zones
1 3 5 7 9 11 13 15 17 19 21 23 25 27 29 31 33 35 37 39 41 43 45 47 49 51 53 55 57 59 61 63 65 67 69 71 73 75 77 79
Segment No.
Tamping Ballast Cleaning
ACCEPTED MANUSCRIPT
AC
C
EP
TE
D
 M
AN
U
SC
R
IP
T
 
16 
 
long terms of the ballast behaviour; (2) they are easy to be derived as the required data are 
obtained from automated inspections (i.e., cost efficient method). 
The new approach takes into account the ballast contamination (as the main indication of 
ballast deterioration) and ballast geometry deformations (as the main indication of ballast 
stability). The efficiency of the new maintenance approach was illustrated by demonstrating its 
application in a railway line. Comparisons of the results, obtained from the applications of the 
conventional and the new maintenance approaches in a railway line, indicate that the new 
proposed algorithm has advantages of providing the suitable timing of tamping, cleaning or 
replacement of the ballast (i.e., more accurate prioritization of ballast maintenance actions). It 
eliminates short interval tamping actions by differentiating the timings of the ballast cleaning 
and tamping. In the new algorithm, the critical zones, for which urgent repair action are 
required, are flagged. This prevents unexpected track failure (such as derailment) and ensures 
the track safety. 
Through more effective allocation of resources and more accurate maintenance planning, the 
new algorithm causes a considerable improvement in the maintenance effectiveness and an 
increase in the life cycle of railway tracks. 
 
7-Refrences 
 
Anbazhagan, P., Dixit, P. N., & Bharatha, T. (2016). Identification of type and degree of railway ballast 
fouling using ground coupled GPR antennas. Journal of Applied Geophysics, 126, 183-190. 
Anbazhagan, P., Lijun, S., Buddhima, I., & Cholachat, R. (2011). Model track studies on fouled ballast 
using ground penetrating radar and multichannel analysis of surface wave. Journal of Applied 
Geophysics, 74(4), 175-184. 
Anderson, P., Cunningham, C., & Barry, D. A. (2002). Efficiency and potential environmental impacts 
of different cleaning agents used on contaminated railway ballast. Land Contamination & 
Reclamation, 10(2), 71-78. 
Auer, F. (2013). Multi-function track recording cars. Rail Technology Review, 53(3/4). 
Berawi, A. R. B. (2013). Improving railway track maintenance using power spectral density (PSD). 
Universidade do Porto (Portugal). 
Berawi, A. R. B., Delgado, R., Calçada, R., & Vale, C. (2010). Evaluating track geometrical quality 
through different methodologies. International Journal of Technology, 1(1), 38-47. 
Brough, M., Stirling, A., Ghataora, G., & Madelin, K. (2003). Evaluation of railway trackbed and 
formation: a case study. NDT & E International, 36(3), 145-156. 
Caetano, L. F., & Teixeira, P. F. (2015). Optimisation model to schedule railway track renewal 
operations: a life-cycle cost approach. Structure and infrastructure engineering, 11(11), 1524-
1536. 
Camargo, L. F. M., Edwards, J. R., & Barkan, C. P. (2011). Emerging condition monitoring 
technologies for railway track components and special trackwork. Paper presented at the 2011 
Joint Rail Conference. 
Clark, M., Gordon, M., Giannopoulos, A., & Forde, M. (2004). Advanced analysis of ground 
penetrating radar signals on railway. Paper presented at the Proceedings of the 7th 
International Conference on Railway engineering, London, UK. 
De Bold, R., O’Connor, G., Morrissey, J., & Forde, M. (2015). Benchmarking large scale GPR 
experiments on railway ballast. Construction and Building Materials, 92, 31-42. 
Fontul, S., Fortunato, E., De Chiara, F., Burrinha, R., & Baldeiras, M. (2016). Railways Track 
Characterization Using Ground Penetrating Radar. Procedia Engineering, 143, 1193-1200. 
Frangopol, D. M., & Liu, M. (2007). Maintenance and management of civil infrastructure based on 
condition, safety, optimization, and life-cycle cost∗. Structure and infrastructure engineering, 
3(1), 29-41. 
ACCEPTED MANUSCRIPT
AC
C
EP
TE
D
 M
AN
U
SC
R
IP
T
 
17 
 
Gallagher, G. P. (1999). Investigation of Railway Trackbed Deterioration using Ground 8 Penetrating 
Rada. MSc Thesis, University of Edinburgh. 
Guler, H. (2014). Prediction of railway track geometry deterioration using artificial neural networks: a 
case study for Turkish state railways. Structure and infrastructure engineering, 10(5), 614-626. 
Hugenschmidt, J. (2000). Railway track inspection using GPR. Journal of Applied Geophysics, 43(2), 
147-155. 
Hyslip, J. P., Smith, S. S., Olhoeft, G. R., & Selig, E. T. (2003). Assessment of railway track 
substructure condition using ground penetrating radar.Paper presented at the 2003 Annual 
Conference of AREMA. 
IMRT. (2005). Railway track superstructure general technical specifications ballast specification (Vol. 
no. 301). Iran: IMRT Ministry Publication. 
Jiménez-Redondo, N., Escriba, S., Benítez, F., Cores, F., & Cáceres, N. (2014). Towards automated 
and cost-efficient track maintenance. Final developments of the ACEM-Rail project. Transport 
Research Arena 2014, 14-17. 
Khakiev, Z., Shapovalov, V., Kruglikov, A., & Yavna, V. (2014). GPR determination of physical 
parameters of railway structural layers. Journal of Applied Geophysics, 106, 139-145. 
Kim, J., Ahn, Y., & Yeo, H. (2016). A comparative study of time-based maintenance and condition-
based maintenance for optimal choice of maintenance policy. Structure and infrastructure 
engineering, 12(12), 1525-1536. 
Li, D., Hyslip, J., Sussmann, T., & Chrismer, S. (2016). Railway geotechnics: Taylor & Francis Group, 
LLC. 
Li, Z., Lei, X., & Gao, L. (2016). A Novel Approach to Assessing Railway Track Quality Based on 
Ensemble Empirical Mode Decomposition. Paper presented at the International Conference on 
Transportation and Development 2016. 
Lichtberger, B. (2005). Track compendium: formation, permanent way, maintenance, economics: 
Eurailpress. 
Madejski, J., & Grabczyk, J. (2002). Continuous geometry measurement for diagnostics of tracks and 
switches. Paper presented at the International Conference on Switches: Switch to Delft. 
Manacorda, G., & Simi, A. (2012). Non-destructive inspection and characterization of track bed with 
microwaves. Paper presented at the Ground Penetrating Radar (GPR), 2012 14th International 
Conference on. 
Misar, H. (2002). IntelligentBallast ManagementWillCutCosts. INTERNATIONAL RAILWAY 
JOURNAL AND RAPID TRANSIT REVIEW, 42(8). 
Navikas, D., Bulevičius, M., & Sivilevičius, H. (2016). Determination and evaluation of railway 
aggregate sub-ballast gradation and other properties variation. Journal of Civil Engineering and 
Management, 22(5), 699-710. 
Nederlof, C., & Dings, P. (2010). Monitoring track condition to improve asset management. UIC WG 
track condition monitoring synthesis report, International Union of Railways, Paris. 
Nimbalkar, S., Indraratna, B., Dash, S. K., & Christie, D. (2012). Improved performance of railway 
ballast under impact loads using shock mats. Journal of Geotechnical and Geoenvironmental 
Engineering, 138(3), 281-294. 
Nurmikolu, A. (2005). Degradation and frost susceptibility of crushed rock aggregates used in 
structural layers of railway track. 
Olhoeft, G. R., & Selig, E. T. (2002). Ground-penetrating radar evaluation of railway track 
substructure conditions. Paper presented at the Ninth International Conference on Ground 
Penetrating Radar (GPR2002). 
Orlando, L., Cardarelli, E., Cercato, M., De Donno, G., & Di Giambattista, L. (2017). Pavement testing 
by integrated geophysical methods: Feasibility, resolution and diagnostic potential. Journal of 
Applied Geophysics, 136, 462-473. 
Pilecki, Z., Isakow, Z., Czarny, R., Pilecka, E., Harba, P., & Barnaś, M. (2017). Capabilities of seismic 
and georadar 2D/3D imaging of shallow subsurface of transport route using the Seismobile 
system. Journal of Applied Geophysics, 143, 31-41. 
Roberts, R., Al-Audi, I., Tutumluer, E., & Boyle, J. (2008). Subsurface Evaluation of Railway Track 
Using Ground Penetrating Radar. 
ACCEPTED MANUSCRIPT
AC
C
EP
TE
D
 M
AN
U
SC
R
IP
T
 
18 
 
Sadeghi, J. (2010). Development of railway track geometry indexes based on statistical distribution of 
geometry data. Journal of Transportation Engineering, 136(8), 693-700. 
Sadeghi, J., & Askarinejad, H. (2010). Development of improved railway track degradation models. 
Structure and infrastructure engineering, 6(6), 675-688. 
Sadeghi, J., & Askarinejad, H. (2011). Development of track condition assessment model based on 
visual inspection. Structure and infrastructure engineering, 7(12), 895-905. 
Santos, V. R. N., & Teixeira, F. L. (2017). Application of time-reversal-based processing techniques to 
enhance detection of GPR targets. Journal of Applied Geophysics, 146, 80-94. 
Scanlan, K. M., Hendry, M. T., & Martin, C. D. (2016). Evaluating the Equivalency Between Track 
Quality Indices and the Minimum Track Geometry Threshold Exceedances Along a Canadian 
Freight Railway. Paper presented at the 2016 Joint Rail Conference. 
Selig, E. T., & Waters, J. M. (1994). Track geotechnology and substructure management: Thomas 
Telford. 
Sharpe, P. (2000). Trackbed investigation. Paper presented at the Journal and report of proceedings-
Permanent Way Institution. 
Tennakoon, N., & Indraratna, B. (2014). Behaviour of clay-fouled ballast under cyclic loading. 
Geotechnique, 64(6), 502-506. 
Uzarski, D. R., Darter, M. I., & Thompson, M. R. (1993). Development of condition indexes for low-
volume railroad trackage. Transportation research record(1381). 
Vale, C., & Ribeiro, I. M. (2014). Railway condition-based maintenance model with stochastic 
deterioration. Journal of Civil Engineering and Management, 20(5), 686-692. 
doi:10.3846/13923730.2013.802711 
Van der Westhuizen, N. J. (2012). The integration of railway condition monitoring technologies to 
establish continuous track asset management improvement. University of Pretoria. 
 
 
ACCEPTED MANUSCRIPT
AC
C
EP
TE
D
 M
AN
U
SC
R
IP
T
 
19 
 
Highlights 
 Two new ballast condition indices were developed. 
 Making use of the new indices, a new ballast maintenance algorithm/approach was 
established. 
 This paper presents a non-contact approach, capable of performing more accurate and 
effective inspection of ballast conditions. 
 Through more effective allocation of resources and more accurate maintenance 
planning, the new algorithm causes a considerable improvement in the maintenance 
effectiveness and an increase in the life cycle of railway tracks. 
ACCEPTED MANUSCRIPT