You are on page 1of 15

This article was downloaded by: [Dalhousie University]

On: 13 November 2013, At: 00:31


Publisher: Taylor & Francis
Informa Ltd Registered in England and Wales Registered Number: 1072954 Registered office: Mortimer House,
37-41 Mortimer Street, London W1T 3JH, UK

Structure and Infrastructure Engineering:


Maintenance, Management, Life-Cycle Design and
Performance
Publication details, including instructions for authors and subscription information:
http://www.tandfonline.com/loi/nsie20

Development of improved railway track degradation


models
a a
Javad Sadeghi & Hossein Askarinejad
a
School of Railway Engineering, Iran University of Science and Technology , Tehran, Iran
Published online: 08 Jul 2008.

To cite this article: Javad Sadeghi & Hossein Askarinejad (2010) Development of improved railway track degradation models,
Structure and Infrastructure Engineering: Maintenance, Management, Life-Cycle Design and Performance, 6:6, 675-688, DOI:
10.1080/15732470801902436

To link to this article: http://dx.doi.org/10.1080/15732470801902436

PLEASE SCROLL DOWN FOR ARTICLE

Taylor & Francis makes every effort to ensure the accuracy of all the information (the “Content”) contained
in the publications on our platform. However, Taylor & Francis, our agents, and our licensors make no
representations or warranties whatsoever as to the accuracy, completeness, or suitability for any purpose of the
Content. Any opinions and views expressed in this publication are the opinions and views of the authors, and
are not the views of or endorsed by Taylor & Francis. The accuracy of the Content should not be relied upon and
should be independently verified with primary sources of information. Taylor and Francis shall not be liable for
any losses, actions, claims, proceedings, demands, costs, expenses, damages, and other liabilities whatsoever
or howsoever caused arising directly or indirectly in connection with, in relation to or arising out of the use of
the Content.
This article may be used for research, teaching, and private study purposes. Any substantial or systematic
reproduction, redistribution, reselling, loan, sub-licensing, systematic supply, or distribution in any
form to anyone is expressly forbidden. Terms & Conditions of access and use can be found at http://
www.tandfonline.com/page/terms-and-conditions
Structure and Infrastructure Engineering
Vol. 6, No. 6, December 2010, 675–688

Development of improved railway track degradation models


Javad Sadeghi* and Hossein Askarinejad
School of Railway Engineering, Iran University of Science and Technology, Tehran, Iran
(Received 23 April 2007; accepted 29 December 2007; published online 8 July 2008)

There is a need among railway track maintenance programmers for a better understanding of the long-term
behaviour of railway track systems and for improved predictive techniques. The use of accurate techniques to predict
railway track conditions increases track safety and maintenance effectiveness. The aim of this research is to improve
current track deterioration modelling techniques using a comprehensive field investigation. Statistical and
engineering approaches are adapted for this research and comprehensive track field data is collected and analysed
over a period of 2 years on approximately 100 km of railway line. Parameters that influence track degradation,
Downloaded by [Dalhousie University] at 00:31 13 November 2013

including loading conditions, track maintenance status and track quality, are investigated, leading to the
development of correlations between the effective parameters and the track degradation. These constructed
correlations are then used to develop a new track degradation model in two formats. One format, based on track
geometry, relies on data obtained from track geometry inspections. The other, based on track structure, relies on
data obtained from visual inspections of the track structural conditions.
Keywords: railway track; degradation models; structure index; geometry index

the current models are taken into consideration, while


1. Introduction maintaining its simplicity. For the development of the
Considerable effort and cost is expended on railway new model, engineering and statistical analyses are
track maintenance. Improvement of maintenance conducted on comprehensive data obtained during the
planning requires an objective measure of track quality field research period. The data is recorded in central
and the track degradation ratio. The prediction and Iran from observations of track performance in a
evaluation of track maintenance costs and the selection 100 km test zone over a 2 year period.
of the most economical maintenance practices is This paper presents:
possible with a track degradation model that takes
into consideration the main variations in traffic, . A review of the track models currently available
structure and maintenance parameters. One of the and a brief discussion of their limitations;
major problems in scheduling and controlling main- . The methodology used for this research and a
tenance arises because a suitably comprehensive model description of the comprehensive field investiga-
to predict the long-tem behaviour of tracks does not tion undertaken; and
exist (Mundrey 2003). Previously developed models . The formulations of the new degradation model.
have some deficiencies that limit their applicability,
practicability, or reliability (Sadeghi 2005). By addres-
sing these deficiencies and limitations, this research is
anticipated to improve current track degradation 2. The current track degradation models
modelling. In the last two decades, several attempts have been
In this research, the effective parameters influen- made to develop a predictive degradation model for
cing the degradation of track systems are investigated, railway tracks. A review of the available literature
track quality is quantified, and the correlations indicates that the main railway degradation models are
between the track degradation coefficient and traffic those developed by Shenton (1985), Sato (1997),
parameters, maintenance factors and track quality are Hamid and Gross (1981), Bing and Gross (1983) and
formulated, allowing the development of an improved Chrismer and Selig (1993).
form of track degradation model. In the improved In the Shenton (1985) degradation model, the main
degradation model, some important factors omitted in controlling factor is the settlement of the ballast.

*Corresponding author. Email: Javad_Sadeghi@iust.ac.ir

ISSN 1573-2479 print/ISSN 1744-8980 online


Ó 2010 Taylor & Francis
DOI: 10.1080/15732470801902436
http://www.informaworld.com
676 J. Sadeghi and H. Askarinejad

Shenton defined six mechanisms of deterioration and, autoregressive techniques to develop predictive equa-
through the evaluation of these mechanisms, indicated tions of the following type:
that the main contribution to track deterioration
comes from the variability of ballast settlement.
Shenton investigated various parameters that influence y ¼ a þ b0 y0 þ b1 x1 þ    þ bm xm ; ð3Þ
ballast settlement and derived a general equation that
quantifies the settlement of ballast: where y is the current TQI, y’ is the previous TQI, a is
an estimated constant term, b0 to bm are the estimated
regression coefficients and x1 to xm are physical
Ae  
S ¼ KS ð0:69 þ 0:028LÞN0:2 þ 2:7  106 N ; ð1Þ parameters. Hamid and Gross indicated that the
20 addition of successive variables can improve the
predictability of the equation. In this model, TQI has
where Ae is the equivalent axle load, N is the total a linear relation with physical parameters. This model
number of axles, L is the lift given by tamping does not consider the track structural conditions and
machines and KS is a factor that is a function of does not yet indicate a procedure for the calculation of
sleeper type and size, ballast type and the subgrade TQI. This model is still under development.
Downloaded by [Dalhousie University] at 00:31 13 November 2013

condition. Shenton’s model does not take into In the Bing and Gross (1983) degradation model,
consideration a number of influencing factors such as TQIs are calculated from the raw track geometry data
the vehicle speed and some of the track components for a track segment. They defined different causal
conditions. A practical and reliable method of parameters in the areas of traffic, track structure and
measuring KS has not been presented. maintenance. Using statistical and engineering ana-
In the Sato (1997) degradation model, track lyses, they presented a general form of the model and
deterioration is quantified by the growth rate of track calibrated it for two test zones. The TQIs chosen for
irregularity over time. Sato considered the coefficient this track deterioration model are the standard
of track deterioration to be in proportion to the deviation of the filtered cross-level and the standard
growth of track irregularity and derived the following deviation of the filtered profile. The model’s general
equation: form is:

     
S ¼ 2:09  103 T0:31 V0:98 M1:3 L0:21 P0:26 ; ð2Þ TQI2 VE a3 RA a4 TQI1P a5
¼ a1 exp ða2 TSSÞ
TQI1 VE RA TQI 1P
 a6  a7  a9
where S is the average growth of track irregularity in a TQI1C BI a8 EMGT
track section (mm/100 days), T is the passing tonnage  ð1 þ FSÞ
TQI 1C BI EMGT
(million tons/year), V is the average running speed
ð4Þ
(km/h), M is the structure factor, L is the influence
factor for jointed rail or continuous welded rail (CWR)
and P is a factor related to the subgrade conditions. In where TQI1 and TQI2 are the initial and the predicted
this equation, the structure factor is expressed as a track quality index in a time period, TSS is the time
function of spring coefficient between rail and tie, since surfacing (months), VE is the equivalent train
effective intermediate mass of tie and ballast, rail speed (km/h), RA is the rail age (years), BI is the
rigidity and the characteristics of the track car. The ballast index, EMGT is equivalent million gross tons
computation of the structure factor is complicated; the and FS is fraction of segment surfaced. The parameters
track spring coefficient is a variable quantity in a1 to a9 are determined by statistical analysis, and VE*,
different conditions and cannot be easily measured or TQIC*, TQIP*, EMGT*, RA* and BI* are the
calculated. arbitrary fixed reference values of each parameter.
In the Hamid and Gross (1981) degradation model, The suffixes C and P refer to cross-level and profile
track quality indices (TQIs) are figures of merit that in TQIs.
objectively quantify the track condition. According to Chrismer and Selig (1993) considered the settle-
Hamid and Gross, TQIs can monitor track degrada- ment of a track as the main controlling factor of track
tion and maintenance operations and display the degradation. According to them, track settlement
condition of large sections of track. They derived arises from the plastic strain of the ballast, sub-ballast,
TQIs from the data collected by automatic track and subgrade. They conducted a test to measure the
geometry cars. TQIs utilise the standard deviations of ballast, sub-ballast, and subgrade contributions rela-
gauge, warp, alignment and unbalanced super- tive to the settlement of a newly constructed track. In
elevation data. Hamid and Gross used linear their model, the total settlement is obtained by
Structure and Infrastructure Engineering 677

multiplying the ballast and sub-ballast strains by their suggested by engineering analyses and that which lends
respective thicknesses and adding this to the subgrade itself to statistical analyses. In this research, the
compression: engineering and statistical analyses rely on field
observations of the performance of a railway track
SL ¼ eN ball  Hball þ eN subball  Hsubball þ dN subg ; ð5Þ main line in central Iran.

where eN is the permanent strain after N load cycles,


dN subg is the subgrade compression after N load cycles 4. Main track degradation parameters
and H is the thickness. The main parameters, influencing the rate of track
As noted above, these models have some limita- degradation, can be categorised into three areas: track
tions that could limit their accuracy and applicability. quality (TQIs), traffic and maintenance (IMRT 2005).
This research seeks to address these limitations and The TQIs are quantities that are used to represent
introduces a new degradation model that remains observed track data, and relate to the ability of the
simple for users, despite the incorporation of most of track to perform its function (AREMA 2006).
the parameters omitted in the previously developed According to the literature, two different track quality
models. indices are commonly used: the track geometry index
Downloaded by [Dalhousie University] at 00:31 13 November 2013

(TGI) and the track structure index (TSI) (Sadeghi and


Akbari 2006). These two indices are not independent
3. Research methodology variables. The TGI is track quality from the perspec-
Track degradation modelling process involves the use tive of the user (similar to IRI in Road Engineering). It
of engineering and/or statistical approaches to produce indicates only the geometry conditions of the track,
an equation that predicts a future track condition by such as profile, twist, gauge and alignment that have a
taking into account all the effective parameters, as well direct influence on track riding comfort. The TGI is
as measured changes. The engineering approach obtained from the statistical analysis of track geometry
consists of establishing, by theory and testing, the data recorded by track geometry recording cars. On
mechanical properties of all the elements that make up the other hand, the TSI represents the structural
the track structure and the railroad vehicles. The conditions of the track, which include the condition
advantage of this approach is that it provides a good of rail, sleeper, fastening systems, subgrade and
engineering understanding of how a track responds to drainage systems. This index refers directly to the
vehicle loading. The disadvantage is that the mechan- actual condition of the track and can indicate the
ical properties of vehicles and track are variable and track’s potential for degradation. The use of the TGI
difficult to quantify. This means that absolute predic- or TSI in the development of track degradation models
tions of track degradation are difficult. The statistical depends on the purpose for establishing such models; a
approach involves the analysis of many observations low value for these two indices indicates different
of actual track performance and the corresponding repair strategies and maintenance requirements.
causal parameters. The track performance, as mea- This research utilises both indices. That is, the
sured by the TQIs or the changes in the TQIs, is the degradation model is developed into two formats. One
dependent variable, and the causal parameters are the takes into consideration the track quality from a
independent variables. Correlation analysis, analysis of structural perspective and the other incorporates the
variance and regression analysis are used to develop track quality from the track geometry perspective
the track degradation models. The advantage of this (riding comfort). In the model developed in this paper,
approach is that because actual observations are used, the TGI and the TSI are considered as effective
an absolute prediction of track performance can parameters. These indices are influenced by the
be given. The disadvantage is that without the qualitative condition of track components, track
engineering understanding, inappropriate model forms geometry condition and any maintenance operations.
may result. Vehicle type, speed, tonnage, axle load and load
In this research, a combined approach is adopted. cycle may contribute to the track’s long-term beha-
First, an engineering analysis is performed in order to viour; however, the main contribution to track
investigate the role of each track and vehicle parameter deterioration comes from the tonnage and the running
in the track degradation. Statistical analyses are then speed (Mundrey 2003). Therefore, the total million
used to construct mathematical expressions for the gross tons passing the track (EMGT) in a time period
relationships between the effective parameters and the (T) and the average running speed (V) are taken into
track degradation ratio, leading to the development of consideration as the most effective traffic parameters.
a predictive model for the track quality conditions. In Any maintenance operations can have an influence on
practice, the model form is a compromise between that the track degradation ratio and, therefore, the time
678 J. Sadeghi and H. Askarinejad

since any major maintenance operation (T) is con- is calculated using the following formula (Mundrey
sidered as the main maintenance parameter. Methods 2003):
by which the track quality indices and traffic and
maintenance parameters can be obtained are discussed 2UI þ TI þ 6AI þ GI
TGI ¼ : ð7Þ
below. 10

In comparison with the other indices, the advantages


4.1. Track geometry parameters of using the TGI are: (1) it provides information on the
There are several indices developed to represent track quality for a continuous length of track rather
the track geometry conditions. They include composite than highlighting isolated low-condition locations; (2)
track record (CTR), track geometry index (TGI) it gives weightings to different parameters according to
and track roughness index (TRI). Among these indices, their effects on the ride index; and (3) it is not overly
the TGI is the more practical and reliable (Sadeghi and affected by minor changes in the track geometry. Due
Akbari 2006). The TGI is used as an indication of the to these advantages, this index was used in the present
quality of the geometry of a railway line in most research.
countries (Mundrey 2003). It is a standard deviation
Downloaded by [Dalhousie University] at 00:31 13 November 2013

based index evolved by Indian Railways. TGI takes


into account the index for different geometry para- 4.2. Track structural parameters
meters, assessing the track condition with respect to The TSI has been designed to provide an objective
the range defined by a best-maintained track on one and quantitative means for facility condition assess-
side and a track needing urgent maintenance on the ment while providing for a common language and
other (Mundrey 2003). Four track geometry para- interpretation among users (Uzarski et al. 1993). This
meters are considered in this index. They are twist (TI), index was first developed at the US Army Construc-
alignment (AI), gauge (GI) and unevenness (UI). The tion Engineering Research Laboratories (USCERL)
indices for the parameters are worked out as follows by Uzarski (1993a). During 2000 to 2006, compre-
(Mundrey 2003): hensive research for the establishment of the Iranian
Railway Track Maintenance Management System
  (ITMMS), funded by the Ministry of Road and
SDm  SDn Transportation of Iran, was made under the direction
GI=TI=AI=UI ¼ 100 exp  ; ð6Þ
SDu  SDn of these authors, in which the TSI was considerably
revised. By addressing the limitations of Uzarski’s
where GI, TI, AI and UI are twist, alignment, gauge work, the TSI reliability and practicality were
and unevenness indices respectively, SDm is the improved (Sadeghi 2005). The establishment of this
standard deviation of the measured parameter, SDn index is based on visual inspections of railway lines
represents the standard deviation values for a newly and the identification of the observable distresses.
laid track and SDu the standard deviation values for a This index is constructed by recording actual track
track needing urgent maintenance. The standard structural defects and applying statistical methods. In
deviation values from (Sadeghi 2005) are shown in the development of this index, the railway track
Table 1. system is considered as a system encompassing three
Considering the effect of each parameter on the ride different groups:
index, different weightings have been assigned to the
geometry parameters. Hence, the track geometry index . The rail group, consisting of rails, fastenings,
joints and rail pads;
. The sleeper group consisting of sleepers; and
. The ballast group consisting of ballast, sub-
Table 1. Standard deviation (SD) values (Sadeghi 2005). ballast, subgrade and drainage systems.
SD SD values for urgent
values, maintenance (mm) The possible defects defined for each group are detailed
Chord newly in Table 2.
length laid track Speed Speed To construct this index, one must make a visual
Parameters (m) (mm) 4105 km/h 5105 km/h
inspection of the line and keep a record of every
Unevenness 9.6 2.50 6.2 7.2 observed defect noting type and severity. The method
Twist 3.6 1.75 3.8 4.2 of recording data should be as established by the
Gauge - 1.00 3.6 3.6 USCERL (Uzarski 1993b) and the ITMMS (Sadeghi
Alignment 7.2 1.50 3.0 3.0
2005). When the recording is completed, an index can
Structure and Infrastructure Engineering 679

Table 2. Distresses in TSI. where CI (low), CI (mid) and CI (high) are the lowest,
Group Distress types medium and highest amounts for RCI, SCI and BCI,
respectively. More details can be found in Sadeghi
Rail R1: rail defects, R2: joint defects, R3: (2005).
fastening defects, R4: base plate (pad)
defects, R5: gauge rod defects, R6: rail
anchor defect
4.3. Traffic and maintenance parameters
Sleeper S1: single defective sleeper, S2: defective
sleeper cluster, S3: defective sleeper clus- According to the literature, vehicle type, speed,
ter that includes one joint sleeper, S4: tonnage, axle load and load cycle may make contribu-
adjacent defective sleeper cluster, S5: all tions to the track long-term behaviour; however, based
joint sleepers defective, S6: missing slee-
pers, S7: all joint sleepers missing, S8:
on previous experience (Mundrey 2003), the main
improperly positioned sleeper contribution to track deterioration comes from the
Ballast B1: dirty (fouled) ballast, B2: vegetation
tonnage and the running speed. In this research, the
growth, B3: settlement of ballast or total million gross tons passing the track (EMGT) in a
subgrade, B4: hanging sleepers at bridge time period (T), and the average running speed (V) are
approach, B5: Centre bound track, B6: taken into consideration for the development of
pumping sleepers, B7: alignment devia-
Downloaded by [Dalhousie University] at 00:31 13 November 2013

tion, B8: insufficient crib/shoulder ballast, degradation models.


B9: erosion of ballast, B10: inadequate Any maintenance operation such as machine
trackside drainage, B11: inadequate water smoothing, hand smoothing, re-gauging, tie renewal,
flow through drainage structures rail renewal, surfacing, etc. affects the quality condi-
tions of the track and, in turn, affects the deterioration
coefficient (the ratio of the future condition of the
be constructed for each component group of rail, track over its present condition). Therefore, a reliable
sleeper and ballast using the following model devel- deterioration coefficient over a time period can be
oped in USCERL and the ITMMS. This index varies obtained when no maintenance operation is allowed in
from 0 for a line with a considerable number of defects that period. In this research, the time since any main
to 100 for a line without any defects: maintenance operation (T) is considered as the main-
tenance parameter (i.e. the time in which no major
X
p X
mi repair is allowed).
ðBCI; SCI; RCIÞ ¼ 100  aðTi ; Sj ; Dij ÞFðt; dÞ;
i¼1 j¼1

ð8Þ 5. Influences of effective parameters on track


degradation
where BCI, SCI and RCI are ballast, sleeper and rail In the previous sections, the main parameters influen-
condition indices, respectively, a( ) is the deduct cing the track degradation are discussed. Now, to
weighting value depending on distress type (Ti), derive a realistic track deterioration model, a statistical
severity level (S) and distress density (Dij), i is the analysis is needed to predict how the degradation
counter for distress types, j is the counter for severity coefficient changes as a function of TQI, passing
levels, p is the total number of distress types for the tonnage, average running speed and time since
component group under consideration, mi is the surfacing. The statistical analysis relies on a series of
number of severity levels for the ith distress type and field investigations into the actual performance of test
F( ) is the adjustment factor for multiple distresses that zones in central Iran. These observations were per-
vary with the total summed deduct value (t) and the formed from July 2004 to September 2006 on
number of individual deducts over an established approximately 100 km of main railway line. A sche-
minimum value (d). The amount of a(T,S,D) and matic view of the test zone is presented in Figure 1. The
F(t,d) can be obtained using the equations developed in test zone includes eight blocks. Each block is between
the ITMMS (Sadeghi 2005). The development of these two subsequent stations, with an average length of
equations is based on engineering judgements followed 12 km. Each block is divided into several segments
by statistical analyses of the data obtained from the with an average length of 0.6 km (they vary from 0.2 to
railway field research. TSI is obtained by the combina- 1.0 km). Depending on the properties of the sub-
tion of these three indices as: structure of the track, segments are categorised into
seven types, as indicated in Table 3. The characteristics
of the infrastructure in the blocks are the same. They
TSI ¼ 0:5CIðlowÞ þ 0:35CIðmidÞ þ 0:15CIðhighÞ;
have pre-stressed concrete sleepers (type B70), long
ð9Þ welding UIC60 rail, 30 cm ballast (crushed stone with
680 J. Sadeghi and H. Askarinejad
Downloaded by [Dalhousie University] at 00:31 13 November 2013

Figure 1. Schematic view of the location of the test zone.

Table 3. Blocks characteristics in the test zone.

Number of segments
Length
Number Commencing and of block Straight Curve- Curve-
of block ending stations (km) line Curve Bridge Tunnel bridge tunnel Turnout
1 A B 12 9 8 4 2 1 2 1
2 B C 14 11 7 3 1 1 3 2
3 C D 10 6 7 3 2 1 2 2
4 D E 12 9 6 2 3 2 2 2
5 E F 13 10 4 2 2 1 4 2
6 F G 10 8 5 1 2 3 3 2
7 G H 14 9 10 2 3 2 2 2
8 H I 12 10 7 1 2 3 2 1

diameter of 40 to 50 mm), and the Wossloh fastening TQI1) due to the changes in the remaining parameters
system. are recorded.
In order to investigate the role of each influencing A bench mark (or reference value) is made for the
parameter (i.e. TQI, vehicle speed, equivalent million properties of those parameters kept unchanged (see
gross tons and time period) on the track degradation Table 4). The initial TGI is varied from 100 to 61, the
coefficient, a comprehensive field investigation under initial TSI from 100 to 64, the train speed from 30 to
the support of the Ministry of Road and Transporta- 100 km/h, the total million gross tons from 3 to 18 and
tion of Iran was made. This investigation took more the time period from 0 to 12 months. In all cases, the
than 2 years and the changes to the track degradation test zone is set up based on Table 4. Then, the changes
due to the variations in each parameter were studied. in the track degradation coefficient are recorded while
Parametric analyses were made to derive correlations changing the conditions of one parameter and keeping
between track degradation and the effective para- the conditions of the other parameters unchanged.
meters, using the data recorded in the field. To conduct The quality indices (TGI and TSI) are computed
these analyses, the conditions of all parameters, except based on data collected from the visual and automated
the conditions of one parameter, are kept constant and inspections. The geometric data, consisting of gauge
the changes in the track degradation coefficient (TQI2/ (track plane), profile (longitudinal vertical plane),
Structure and Infrastructure Engineering 681

Table 4. Referenced values, considered for the effective


parameters in these field investigations.

Equivalent
million
Effective Initial Initial Train gross Time
parameters TGI TSI speed tons period
Referenced 100 100 30 km/h1 3 1 month
value

alignment (horizontal plane), cross-level, and twist


(transverse vertical plane) are recorded by an EM120
recording car. The TSI is calculated based on the
model detailed in x4. The results obtained for each
parameter are discussed in the following sections.
Downloaded by [Dalhousie University] at 00:31 13 November 2013

5.1. Train/freight speed


In order to investigate the influence of the train speed
on the degradation coefficient, the initial TGI, initial
TSI and EMGT are set according to Table 4. The
EMGT is kept constant throughout the test zone.
After setting the site, trains are sent on the blocks with
a starting speed of 30 km/h in the first block to 100 km/
h in the last one (i.e. there is a 10 km/h increase in the
train speed at each intermediate station). After a
period of 1 month, the amounts of TGI and TSI for all
the block segments are calculated. By dividing the new
TQIs by the initial TQIs, the degradation coefficient is Figure 2. Train speed versus: (a) geometric degradation
obtained. coefficient (initial TGI ¼ 100), and (b) structural degradation
The results obtained are graphically presented in coefficient (initial TSI ¼ 100), for EMGT ¼ 3 and time
period ¼ 1 month.
Figure 2. Since each block had more than one segment
of each type, an average is made for the results
obtained for the segments with the same type. In this
way, possible errors are minimized. Due to some limitations in the tests, the increase of
The results obtained indicate an exponential form train speeds for more than 100 km/h was not possible;
of relationship between track degradation coefficient however, from the results obtained (presented in
and train speed. Using the method of least squared Figure 2), a prediction (approximation) of the degra-
errors, the influences of train speed on the track dation coefficient for the train speed of more that 100
degradation coefficient can be expressed as: km/h can be made using Equations (10) and (11).

C1 ¼ a1 exp ðb1 VÞ ð10Þ 5.2. Train/freight loads


For the investigation of the influence of the train loads
and on the degradation coefficient, the initial TGI and TSI
and train speed are set up according to Table 4 for all
C01 ¼ a01 exp ðb01 VÞ; ð11Þ the blocks. Trains are sent on the blocks with a
constant speed of 30 km/h for a period of 1 month.
where C1 and C01 are the degradation coefficients The amount of EMGT from one bock to another is
representing the influences of changes in the train changed, starting from 3 for the first block to 18 for the
speed on the track degradation from the geometric last one. In order to vary the EMGT from one block to
perspective and the structural perspective respectively the next one, the trains passing the block are reduced
and a1, b1, a01 and b01 are constant coefficients, as at each station by directing them to a different
tabulated in Table 5. direction through turn-outs. After a period of 1 month,
682 J. Sadeghi and H. Askarinejad

Table 5. The values of a1, b1, a01 and b01 for different types of lines.

Line type Straight line Tunnel Curve Curve-tunnel Turnout Bridge Curve-bridge
a1 1.0121 1.0121 1.0151 1.0151 1.0151 1.0151 1.0182
b1 70.0004 70.0004 70.0005 70.0005 70.0005 70.0005 70.0006
a01 1.0090 1.0090 1.0090 1.0121 1.0121 1.0121 1.0121
b01 70.0003 70.0003 70.0003 70.0004 70.0004 70.0004 70.0004

the amounts of TGI and TSI for all the block segments
are calculated. The degradation coefficient is calculated
by dividing the new TQIs by the initial TQIs. The
results obtained are graphically presented in Figure 3.
As in the last section, since each block had more than
one segment of each type, an averaging method is used
Downloaded by [Dalhousie University] at 00:31 13 November 2013

to minimize possible errors.


As for the velocity, the results obtained indicate an
exponential form of relationship between the track
degradation coefficient and the equivalent million
gross tons. Using the method of least squared errors,
the influences of the EMGT on track degradation
coefficient can be expressed as:

C2 ¼ a2 exp ðb2 EMGTÞ ð12Þ

and

C02 ¼ a02 exp ðb02 EMGTÞ; ð13Þ

where C2 and C02 are the degradation coefficients


representing the effects of changes in EMGT on the
track degradation from the geometric perspective
and the structural perspective, respectively and a2, b2,
a02 and b02 are constant coefficients as tabulated in
Table 6.

5.3. Initial TQIs


To investigate how the current track quality influences
the track degradation coefficient, the field (eight Figure 3. Equivalent million gross tons versus: (a)
blocks) is left free for random maintenance operations geometric degradation coefficient (initial TGI ¼ 100), and
(b) structural degradation coefficient (initial TSI ¼ 100), for
for a 6 month period. Replacement of defective track train speed ¼ 30 km/h and time period ¼ 1 month.
components is also allowed. During this period, trains
are left free to pass with different loading conditions.
At the end of this period of time, all the operations are
stopped and the TQI values for all the segments are graphically presented in Figure 4. Again, an averaging
recorded. Taking the recorded values for TQIs as method is used to minimize the possible errors. As
initial TQIs for the segments, trains are sent on the indicated in Figure 4, reduction of the initial TQI
blocks, with the train speed and EMGT as in Table 4. causes a substantial reduction in the degradation
The train speed and the EMGT are kept unchanged for coefficient.
a period of 1 month and, at the end, the amounts of As for the velocity and the EMGT, the results
TGI and TSI for all the block segments are calculated. obtained indicate an exponential form of relationship
By dividing the new TQIs by the initial TQIs, the between the track degradation coefficient and the
degradation coefficient is obtained. The results are initial TQIs. Using the method of least squared errors,
Structure and Infrastructure Engineering 683

Table 6. The values of a2, b2, a02 and b02 for different types of lines.

Line type Straight line Tunnel Curve Curve-tunnel Turnout Bridge Curve-bridge
a2 1.0093 1.0099 1.0108 1.0112 1.0121 1.0127 1.0142
b2 70.0031 70.0033 70.0036 70.0037 70.004 70.0042 70.0047
a02 1.0057 1.0060 1.0069 1.0075 1.0078 1.0087 1.0093
b02 70.0019 70.002 70.0023 70.0025 70.0026 70.0029 70.0031

and initial TSI on the track degradation, respectively


and a3, b3, a03 and b03 are constant coefficients as
tabulated in Table 7.
The results obtained from the parametric analyses
indicate that the influence of initial TQI on the
degradation coefficient is more than that of other
Downloaded by [Dalhousie University] at 00:31 13 November 2013

influencing parameters. The impact of changes in TQIs


on the degradation coefficient is more substantial when
the initial TQI is less than 70. Results indicate that an
increase of train speed from 30 to 100 km/h in a
straight line with a constant passing tonnage causes a
2.8% reduction in the degradation coefficient, while an
increase in equivalent million gross tons (EMGT) from
3 to 18 for a similar line with a constant train speed led
to a 4.5% reduction in the degradation coefficient.
That is, among the traffic parameters, the equivalent
million gross tons passing the track is the most
influential parameter on the track degradation. There
are 2.8% and 2% reductions in the geometry
degradation coefficient (GDC) and the structural
degradation coefficient (SDC), respectively, when
increasing the speed from 30 to 100 km/h in a period
of 1 month for the straight line segments. Also, an
increase of EMGT from 3 to 18 results in a 4.5% and
2.8% reduction in the GDC and the SDC, respectively.
This indicates that the sensitivity of the GDC with
respect to changes in effective parameters is larger than
that of the SDC. The results obtained from the
Figure 4. (a) Initial TGI versus geometric degradation sensitivity analyses indicate that there are considerable
coefficient, and (b) initial TSI versus structural degradation differences in the degradation coefficients obtained
coefficient, for train speed ¼ 30 km/h, time period ¼ 1 from different types of segments (straight line, curves,
month and EMGT ¼ 3. tunnels and bridges, etc.). For example, the increase of
train speed from 30 to 100 km/h in a straight line with
a constant passing tonnage causes a 2.8% reduction in
the influences of initial TQIs on track degradation the degradation coefficient, while the same conditions
coefficient can be expressed as: resulted in 4.1% and 3.4% reductions in the degrada-
tion coefficient for curve-bridge segments and bridge
C3 ¼ a3 exp ðb3 TGI1 Þ ð14Þ lines, respectively. That is, degradation coefficients for
lines in the turnouts, bridges and curve-bridges in
and comparison with the other types of segments are more
sensitive to changes in the effective parameters.
C03 ¼ a03 exp ðb03 TSI1 Þ; ð15Þ In order to achieve the most accurate approxima-
tion of track degradation coefficient, all the possible
where C3 and C03 are the degradation coefficients mathematical forms for the correlation between the
representing the impact of change in the initial TGI results obtained for the track degradation coefficient
684 J. Sadeghi and H. Askarinejad

Table 7. The values of a3, b3, a03 and b03 for different types of lines.

Line type Straight line Tunnel Curve Curve-tunnel Turnout Bridge Curve-bridge
a3 0.8521 0.8521 0.8437 0.8353 0.8269 0.8187 0.8025
b3 0.0016 0.0016 0.0017 0.0018 0.0019 0.0020 0.0022
a03 0.8958 0.8958 0.8869 0.8869 0.8693 0.8607 0.8521
b03 0.0011 0.0011 0.0012 0.0012 0.0014 0.0015 0.0016

and the train speed, EMGT and the initial TQIs are equivalent million gross tons, and initial TQIs, respec-
examined (applying all of the curve-fitting options in tively. To obtain f(T) and f 0 (T), the track and loading
Excel program). As presented in Equations (10) to conditions are set up according to Table 4 and kept
(15), the exponential form was found to be the most constant for a period of 12 months. Then, the TQIs are
suitable approximation, based on the method of least measured every 2 months. The results are graphically
squared errors. The same form of correlation (approx- presented in Figure 5.
imation) has been also recorded by others (e.g. From the results obtained, the most suitable form
Downloaded by [Dalhousie University] at 00:31 13 November 2013

Hoshyari 2005). of mathematical expressions for f(T) and f 0 (T) are


achieved. That is, all the possible forms of approxima-
tion of the geometric and structural based degradation
6. Track degradation model coefficients (f and f 0 ) are examined (applying all of the
The influences of the main effective parameters on the curve-fitting options in Excel program) and, as a result,
track degradation were formulated in the previous a polynomial form (of order 4) was found to be the
section. In order to construct a track degradation most suitable approximation, based on the method of
model, a mathematical expression that indicates the least squared errors. The author has also demonstrated
future conditions of the track should be established, the suitability of such correlation (approximation)
taking into consideration the roles of the track’s main elsewhere (Sadeghi and Fathali 2007). These are:
influencing parameters.
Development of the degradation model in this fðTÞ ¼ l1 T 4 þ l2 T 3 þ l3 T 2 þ l4 T þ 1 ð18Þ
research is based on the construction of a correlation
between the track degradation coefficient and time. and
First, a correlation is obtained between the track
degradation coefficient and the time where all the f 0 ðTÞ ¼ l01 T 4 þ l02 T 3 þ l03 T 2 þ l04 T þ 1; ð19Þ
effective parameters have constant referenced values
according to Table 4; then, by incorporating the where f(T) and f 0 (T) are the geometric and structural
relationships between the degradation coefficient and based degradation coefficients, T is the time and l1
the main track effective parameters, the influences of to l4 and l01 to l04 are constant coefficients as tabulated
these parameters on the track degradation are taken in Table 8.
into account. The mathematical expression of the Substituting Equations (10) to (15), (18) and (19)
model is: into Equations (16) and (17) results in the degradation
models:
TGI2
¼ C1 C2 C3 fðTÞ ð16Þ
TGI1 TGI2
¼ a4 exp ðb1 V þ b2 EMGT þ b3 TGI1 Þ
TGI1
and
 ½l1 T 4 þ l2 T 3 þ l3 T 2 þ l4 T þ 1 ð20Þ
TSI2
¼ C 01 C 02 C 03 f 0 ðTÞ; ð17Þ
TSI1 and

where TGI2, TGI1, TSI2 and TSI1 are the future track TSI2
geometry index, the present track geometry index, the ¼ a04 exp ðb01 V þ b02 EMGT þ b03 TSI1 Þ
TSI1
future track structure index and the present track
structure index, respectively, f(T) and f 0 (T) are the  ½l01 T 4 þ l02 T 3 þ l03 T 2 þ l04 T þ 1; ð21Þ
geometric and structural based degradation coefficients
as a function of time, C1 to C3 and C01 to C03 are the where a4 and a04 are obtained by multiplying a1 to a3
coefficients that are functions of the vehicle speed, and a01 to a03 , respectively.
Structure and Infrastructure Engineering 685

the track structural perspective (based on track visual


7. Correlation between track degradation models inspections). Geometry inspections are made using
In this research, the track degradation model is automatic inspection cars, such as the EM120, which
developed in two formats. One indicates the track can run with a speed of more than 80 km/h, while
degradation from the track geometry perspective visual inspections are made by inspectors, which is a
(based on inspections of the geometry of the track), time consuming process. Some railway industries
and the other one represents track degradation from might have difficulty maintaining a program of
regular structural inspections and are therefore
limited to studying track degradation behaviour
from the geometry aspect only. However, having a
good indication of track behaviour from both
perspectives (geometry and structure) provides the
possibility of obtaining the best method of program-
ming and management of the track maintenance and
repairs. In particular, information on the track
structural conditions will help identify the track
Downloaded by [Dalhousie University] at 00:31 13 November 2013

defects and the causes of the recorded track low


quality conditions (obtained from automated geome-
try inspections) and, therefore, provide the ability to
take a better assessment of the track components
repairs.
If the initial TGI and the initial TSI are the same,
or of negligible difference, an estimation of both the
future TSI and the future TGI can be made if the
initial value for any of the TSIs or TGIs is available.
To draw a mathematical expression for this estima-
tion, correlations between the ratio of TSI2/TGI2 and
the effective parameters are made using the results
obtained from the field. These are graphically
presented in Figure 6.
Using the results presented in Figure 6, a model
representing the correlation between TSI2 and TGI2 is
developed as:

TSI2 ¼ Z1 Z2 Z3 Z4 TGI2 ; ð22Þ

where Z1 to Z4 are the factors representing the


influences of the train speed (V), equivalent million
Figure 5. Time versus: (a) geometric degradation coefficient
(initial TGI ¼ 100), and (b) structural degradation coefficient gross tons (EMGT), initial (TGI1), and time (T) on
(initial TSI ¼ 100), for train speed ¼ 30 km/h and TSI2/TGI2. Obtaining linear correlations between
EMGT ¼ 3. the ratio of TGI2/TSI2 and the influencing

Table 8. The values of l1 to l4 and l01 to l04 for different types of lines.

Line type l1 l2 l3 l4 l01 l02 l03 l04


Straight line 561076 70.0002 0.0022 70.0103 561076 70.0002 0.0022 70.0102
Tunnel 861076 70.0003 0.0027 70.0115 661076 70.0002 0.0023 70.0109
Curve 161075 70.0003 0.003 70.0123 661076 70.0002 0.0025 70.0115
Curve-tunnel 161075 70.0004 0.0034 70.0134 861076 70.0003 0.0029 70.0127
Turnout 261075 70.0005 0.0040 70.0148 961076 70.0003 0.0032 70.0136
Bridge 161075 70.0004 0.0035 70.0139 761076 70.0003 0.0029 70.0130
Curve-bridge 261075 70.0005 0.0044 70.0162 961076 70.0003 0.0032 70.0141
686 J. Sadeghi and H. Askarinejad
Downloaded by [Dalhousie University] at 00:31 13 November 2013

Figure 6. (a) Train speed versus TSI2/TGI2 (initial TGI ¼ 100, EMGT ¼ 3), (b) equivalent million gross tons versus TSI2/
TGI2 (initial TGI ¼ 100, train speed ¼ 30 km/h), (c) initial TGI versus TSI2/TGI2 (EMGT ¼ 3, train speed ¼ 30 km/h), and
(d) time versus TSI2/TGI2 (initial TGI ¼ 100, EMGT ¼ 3, train speed ¼ 30 km/h), all for different types of lines and time
period ¼ 1 month.

Table 9. The values of k1 to k8 for different types of lines.

Line type k1 k2 k3 k4 k5 k6 k7 k8
Straight line 0.0002 0.9940 0.0012 0.9964 70.0006 1.0604 0.0026 0.9978
Tunnel 0.0001 0.9970 0.0013 0.9961 70.0006 1.0604 0.0023 0.9980
Curve 0.0001 0.9970 0.0013 0.9961 70.0007 1.0706 0.0024 0.9992
Curve-tunnel 0.0002 0.9940 0.0012 0.9964 70.0007 1.0706 0.0025 1.0023
Turnout 0.0001 0.9970 0.0014 0.9958 70.0007 1.0706 0.0026 1.0024
Bridge 0.0001 0.9970 0.0013 0.9961 70.0007 1.0706 0.0024 1.0069
Curve-bridge 0.0001 0.9970 0.0016 0.9952 70.0007 1.0706 0.0040 1.0016

parameters, the following expressions are obtained


for Z1 to Z4: 8. Conclusions
Improvements in the current track deterioration
Z1 ¼ k 1 V þ k 2 ; ð23Þ modelling are attempted in this research. They are
made through an engineering and statistical analyses
Z2 ¼ k3 EMGT þ k4 ; ð24Þ of the results obtained from more than 2 years of field
investigations on 100 km of a main railway line in
Z3 ¼ k5 TGI1 þ k6 ð25Þ central Iran. The factors influencing the track
degradation coefficient (the ratio of future track
and quality to the present one) are investigated. The
Z4 ¼ k7 T þ k8 ; ð26Þ analyses of the data obtained from the field indicate
that among traffic, maintenance and track condition
where k1 to k8 are the constant coefficients as tabu- parameters, the initial track quality, the train speed,
lated in Table 9. the total load passing the track and the time since any
Structure and Infrastructure Engineering 687

main track maintenance operations, are found to The degradation prediction from the proposed
have the greatest effects on the track degradation models can be utilised for future maintenance. That
coefficient (DC). is, one can schedule the maintenance and establish a
The research demonstrates that the influence of timetable for future maintenance activities using the
initial TQI on the rate of degradation is more than that degradation models developed in this research. This
of the other influencing parameters, particularly when enables us to predict the required maintenance time
the initial TQI is less than 70. Also, the results resources and, in turn, predict the required budgets for
presented in Figures 2 and 3 indicate that the EMGT future maintenance (i.e. the financial maintenances
has more impact on the track degradation in compar- programming).
ison with the train speed. Furthermore, a comparison More research is needed for both further evaluation
of the results demonstrates that the sensitivity of the of the accuracy of these models and also the
geometry degradation coefficient to the effective development of analytical techniques for maintenance
parameters is larger than that of the structural planning and repair management. Further research in
degradation coefficient. In other words, the geometry this area is in progress.
conditions of the track have a higher rate of degrada-
tion compared with the structural condition of the Acknowledgements
Downloaded by [Dalhousie University] at 00:31 13 November 2013

track. This research was not possible without the comprehensive


In this research, mathematical expressions are financial and technical supports of the Ministry of Road
developed for the correlations between the main and Transportation of Iran. The authors deeply acknowledge
and appreciate their vital and endless support. This
effective parameters (including loading and track appreciation is extended to all hardworking site engineers
maintenance conditions, as well as track quality) and and technicians who carried out all the site works eagerly and
the track degradation coefficient (DC). An equation carefully.
representing the changes in DC in response to time is
also made using the results obtained from the
observations of track behaviour over a period of 1 References
year. This equation is further improved, leading to the American Railway Association and Maintenance Way
development of a track deterioration model by (AREMA), 2006. Track measuring systems. In: Manual
for railway engineering. Washington DC, USA: AREMA
incorporating the constructed correlations between publishing service, 11–17.
the track degradation coefficients and track quality, Bing, A.J. and Gross, A., 1983. Development of railway
loading and maintenance conditions. track degradation models. Transportation Research
The degradation model developed in this research Board, 939, 27–31.
is expressed in two formats. One is based on data Chrismer, S. and Selig, E.T., 1993. Computer model for
ballast maintenance planning. In: Proceedings of 5th
obtained from inspections of the track geometry, and international heavy haul railway conference, Beijing, 223–
it indicates how the track geometry deteriorates with 227.
respect to time. The model in the second format is Hamid, A. and Gross, A., 1981. Track-quality indices and
based on visual inspections of the structural condi- track degradation models for maintenance-of-way plan-
tions of the track components. It indicates how the ning. Transportation Research Board, 802, 2–8.
Hoshyari, I., 2005. Investigation on railway track deteriora-
track structural defects grow with respect to time. tion ratio from structural point of view. Isfahan University
Due to the limitations of some railway industries in of Technology, Iran: IUT Publication Service, Research
achieving both inspections, a correlation between Report No. 234-F11, 34–35.
these two model formats is investigated, leading to Iran Ministry of Roads and Transportation (IMRT), 2005.
the development of a formulation for the correlation Railway track superstructure general technical specifica-
tions. Standard no. 301. Iran: IMRT Ministry Publication
between the TGI2 (future track geometry index) and Service, 9–12.
the TSI2 (future track structure index). Therefore, Mundrey, J.S., 2003. Railway track engineering. New Delhi,
by having the present TGI, one can estimate both India: McGraw–Hill.
the TSI2 and the TGI2. As the behaviour of the Sadeghi, J., 2005. Algorithm of Iranian track maintenance
track in different track segments is different, the management system. Iran University of Science and
Technology, Tehran, Iran: IUST Publishing Service,
degradation model is separately developed for curves, Report No. F11/19016, 18–23.
turn-outs, straight lines, tunnel lines and bridges lines. Sadeghi, J. and Akbari, B., 2006. Field investigation on effect
A comparison of the degradation of these lines of railway track geometry parameters on rail wear.
indicates that tracks in bridges, curve-bridges and International journal of Zhejiang University, Science A, 11
turnouts deteriorate with a higher rate in comparison (7), 1846–1855.
Sadeghi, J. and Fathali, M., 2007. Deterioration analysis
with other types, and therefore need more of flexible pavements under overweight vehicles.
attention, especially in maintenance and inspection Journal of transportation engineering, ASCE, 133 (11),
scheduling. 625–633.
688 J. Sadeghi and H. Askarinejad

Sato, Y., 1997. Optimum track structure considering dete- Uzarski, D.R., 1993b. Condition survey inspection and distress
rioration in ballasted track. In: Proceedings of 6th manual. Washington DC, USA: USACERL, Technical
international heavy haul conference, Cape Town, Report No. FM-93/14.
SouthAfrica, 576–590. Uzarski, D.R., Brown, D.G., Harris, R.W., and Plotkin,
Shenton, M.J., 1985. Ballast deformation and track dete- D.E., 1993. Maintenance management of US army rail-
rioration. Track Technology, Tomas Telford Ltd, 253– road networks – the RAILER system. Washington DC,
264. USA: USCERL, Technical Report No. FM-94/01.
Uzarski, D.R., 1993a. Development of condition indexes for
low volume railroad track. Washington DC, USA:
USACERL, Technical Report No. FM-93/13.
Downloaded by [Dalhousie University] at 00:31 13 November 2013

You might also like