Requirements for Big Data Adoption for Railway Asset Management

Abstract

Nowadays, huge amounts of data have been captured along with the day-to-day operation of assets including railway systems. Hence, we have come to the era of big data. The utilization of big data technologies for asset condition information management is becoming indispensable for improving asset management decision making. The vital information such as precursor information collected on failure modes and knowledge that may be available for analysis is hidden within the large extent of data. There are analysis tools incorporated with techniques such as multiple regression analysis and machine learning that are facilitated by the availability of big data. Therefore, the utilization of big data technologies for asset condition information management is becoming indispensable for improving asset management decision making. This paper provides a review of the requirements and challenges for big data analytics applications to railway asset management. The review focuses on railway asset data collection, data management, data applications with the implementation of Blockchain technology as well as big data analytics technologies. The need for, and the importance of big data analytics in railway asset management; and the requirement for the asset condition data collection in the railway industry are highlighted. Research challenges in railway asset management via application of big data analytics are identified and the future research directions are presented.

Full text

This work is licensed under a Creative Commons Attribution 4.0 License. For more information, see https://creativecommons.org/licenses/by/4.0/.
This article has been accepted for publication in a future issue of this journal, but has not been fully edited. Content may change prior to final publication. Citation information: DOI
10.1109/ACCESS.2020.2967436, IEEE Access
VOLUME 8, 2020 1
Date of publication xxxx 00, 0000, date of current version xxxx 00, 0000.
Digital Object Identifier 10.1109/ACCESS.2017.Doi Number
Requirements for Big Data Adoption for
Railway Asset Management
P. McMahon, Member, IEEE, T. Zhang, Member, IEEE, and R. Dwight, Member, IEEE
1Faculty of Engineering and Information Sciences, University of Wollongong, Wo llongong, NSW 2522 Aust ralia
Corresponding author: T. Zhang (e-mail: tieling@uow.edu.au).
ABSTRACT Nowadays, huge amounts of data have been captured along with the day-to-day operation of
assets including railway systems. Hence, we have come to the era of big data. The utilization of big data
technologies for asset condition information management is becoming indispensable for improving asset
management decision making. The vital information such as precursor information collected on failure
modes and knowledge that may be available for analysis is hidden within the large extent of data. There are
analysis tools incorporated with techniques such as multiple regression analysis and machine learning that
are facilitated by the availability of big data. Therefore, the utilization of big data technologies for asset
condition information management is becoming indispensable for improving asset management decision
making.
This paper provides a review of the requirements and challenges for big data analytics applications to
railway asset management. The review focuses on railway asset data collection, data management, data
applications with the implementation of Blockchain technology as well as big data analytics technologies.
The need for, and the importance of big data analytics in railway asset management; and the requirement
for the asset condition data collection in the railway industry are highlighted. Research challenges in
railway asset management via application of big data analytics are identified and the future research
directions are presented.
INDEX TERMS Big data analytics, asset management, railway, data management, Blockchain
I. INTRODUCTION
The term of ‘Big Data’ was coined by a few pioneer
researchers and scientists in later 1990s ([1], [2], [3]) since
the first proposal of World Wide Web (WWW) was
invented by Berners-Lee in 1989 [4] and its wide
application started in 1993 [5]. Since then, the global data
volume has been growing extremely fast year by year. ‘Big
Data’ refers to “the explosion in the quantity (and
sometimes, quality) of available and potentially relevant
data, largely due to the result of recent and unprecedented
advancements in data recording and storage technology”
[6]. With the advancement of technologies, increasing
numbers of sensors have been implemented in various
industry fields. Therefore, huge amounts of data are
collected in day to day operations in different industry
sectors including railways. Those large volumes of data
bring us great value while with big challenges as well. The
traditional way to handle the data including data storage,
processing and management is obviously incapable of
satisfying the current social and industrial activity needs.
The technologies being capable of dealing with large
volumes of data are therefore required thereby the
appearance of a new term of ‘Big Data Analytics’. ‘Big
Data Analytics’ refers to a process from data collection,
data management to real application by applying available
techniques/technologies applicable to dealing with large
volumes of data, i.e., big data. The discussion and
investigation of related technologies and their applications
have been soon rapidly expanded through almost every
industrial field.
Big data analytics has become a hot research topic in
recent ten years. The number of published articles has been
exponentially increasing especially in the last five years.
These include many review papers discussing the related
technologies and open issues for application of big data and
big data analytics, e.g., [7], [8], [9], [10] and [11] and many
others by focusing on application of big data analytics in
specific industrial fields such as [12] covering agricultural
production, [13] in healthcare, [14] in supply chain
management, [15] in smart manufacturing, and also a very
recent review article on application of big data analytics in
railway transportation systems (RTS) [16]. The key

This work is licensed under a Creative Commons Attribution 4.0 License. For more information, see https://creativecommons.org/licenses/by/4.0/.
This article has been accepted for publication in a future issue of this journal, but has not been fully edited. Content may change prior to final publication. Citation information: DOI
10.1109/ACCESS.2020.2967436, IEEE Access
2
VOLUME 8, 2020
contribution of this review paper given by [16] is the
taxonomy of the selected papers in review with analysis and
discussions on big data sources and by considerations of
applications of big data analytics on maintenance,
operation, and safety of RTS. In view of asset management
life cycle activities and to the best of authors’ knowledge
based on the literature survey, it lacks a comprehensive
review with discussions in detail about the application
advances of big data analytics in railway network area for
railway asset management. From railway system operation
point of view, however, there are huge amounts of data
collected every day. As a result, it has been identified that
this is a typically important area that big data analytics will
show and demonstrate its power. In addition, the migration
towards the digital railway by a number of rail operators
worldwide will increase the uptake of big data analytics
into rail infrastructure asset management [17]. For these
reasons, we think an overall review on big data analytics
and its application status in railway industry is required and
it is also timely needed in order to provide an insight into
the technological development, challenges and gaps for the
railway system operators and researchers.
Railway networks have been one of the largest assets
in most countries and to manage it well has always been a
concern of the railway asset owners and operators. This has
led to the initial idea for asset management within a railway
infrastructure environment which has evolved from a
number of sources including the concept of Total System
Support [18]. As industry support for asset management
systems has developed, standardization processes have
resulted in the development of asset management standards
such as ISO 55000 ~ ISO 55002 [19]. An asset
management system is concerned with the planning and
control of all asset-related activities and their relationships
to ensure that asset performance meets the intended
competitive strategy of the organization. All aspects related
to asset life cycle activities from concept design to disposal
are crucial to the success of an organization. Asset lifecycle
activities can be considered to be both interdisciplinary and
interrelated [20].
Rail assets are capital intensive. They drive a
significant proportion of the organization’s service delivery
costs. Even a small improvement in asset management can
bring a large benefit. The rail industry is concerned with the
condition of its assets and is actively involved in
developing advanced strategies and techniques to maintain
its infrastructure. The motivation for managing the asset
condition and the movement towards condition-based
maintenance has provided an impetus to the adoption of big
data analytics for rail asset management. The challenge that
asset management presents in the rail industry, however,
extends beyond infrastructure and into assets of all classes,
including transportation systems, operations, rolling stock,
services, safety, and security. There is a mix within the
infrastructure of complex linear assets such as track and
electrical overhead wires with discrete assets such as
bridges, switches, and stations [21]. This amalgam of
infrastructure assets needs to be correlated with rolling
stock assets to provide an overview of asset condition
performance including the wheel/rail interface [22], [23].
Asset condition monitoring plays a primary role in
asset management because it provides asset condition
information that enables subsequent activities to be decided
efficiently and effectively. As highly sophisticated
condition monitoring systems produce huge amounts of
data every day, it is obviously impractical or arguably
impossible to handle the data in an alphanumerical form.
Instead, most of the data gathered must be properly
visualized in order for users to gain insight into the actual
behavior of the objects in question. Hence high-quality
visualization of the analyzed data is needed when it comes
to infrastructure management. This has led to the utilization
of evidence-based decision making [24].
In order to provide high-quality decision-support, a
railway asset management system requires a large amount
of data to be available for analysis. This data, however, is
usually contained in various databases and storage systems.
It requires careful data-selection and transfers to a railway
asset management database for infrastructure condition
analysis and work-planning purposes. The data may be in a
structured format (which can include sensor data, real-time
monitoring data, failure codes) or in semi-structured or
unstructured data such as maintenance reports and service
logs. In traditional asset management systems, the
maintenance reports and equipment service logs are
normally stored separately. However, for analysis purposes,
the maintenance reports and equipment service logs may be
reviewed in a combined format. In terms of big data
classification, structured, semi-structured and unstructured
data are often classified under “Content Format” in terms of
their characteristics [8]. For traditional asset management
systems, data selection is known to be very sensitive (and
important) as all the future analyses and subsequent
planning are to be performed based on the data stored in the
asset management database [25]. Hence, the quality and
reliability of the transferred data represent one of the crucial
issues and keys to the success of any railway asset
management system in use. However, data characteristics
are not normally considered in conventional asset
management data models, whereas big data utilizes the data
characteristics to identify heterogeneous formats for
handling [11]. The other issue is the Rule-creation, i.e., the
transfer of the user’s knowledge, standards and regulations
comprising, in fact, the overall maintenance policy into
asset management decision-rules [26]. In very recent years,
the idea of big data analytics within the railway industry
has been introduced [27]. In terms of building transport
network sustainability, the measurement of performance
indicators to show progress towards sustainability can be
challenging in the presence of big data where managing the
amounts of big data can be a significant challenge [28].
Other issues have also been identified in using big data for
predicting the behavior of assets in use.

This work is licensed under a Creative Commons Attribution 4.0 License. For more information, see https://creativecommons.org/licenses/by/4.0/.
This article has been accepted for publication in a future issue of this journal, but has not been fully edited. Content may change prior to final publication. Citation information: DOI
10.1109/ACCESS.2020.2967436, IEEE Access
3
VOLUME 8, 2020
The purpose of this paper is to conduct a thorough
review of big data analytics and its applications in railway
industry in order to identify the gaps in the areas that are
particularly important for asset management and condition
monitoring. More importantly, this paper seeks to identify
cross-disciplinary concepts and opportunities for both rail
asset management practitioners who have traditionally used
small data and are now facing big data challenges, and big
data researchers who are engaged in technological
development for handling big data. It is our view that the
recent progress made in use of big data analytics has the
potential to drive fundamental advances in research in rail
asset performance monitoring and, at the same time, the
knowledge accumulated in transportation research in the
past many decades can guide big data studies to answer
questions that matter to the rail transportation systems.
With the purposes as described above, this paper is
organized as follows: Section II gives a brief introduction to
the methodology utilized in this review; Section III provides
background knowledge about big data and big data analytics
in railway asset management; Section IV gives an overview
of railway asset data collection requirements in big data
environments; Section V discusses the big data analytics
technologies available for implementation where the
challenges and technical gaps are also highlighted; Section
VI presents a brief overview of big data analytics
applications to railway infrastructure where challenges in the
applications are indicated, and this paper is concluded as
given in Section VII.
II. METHODOLOGY
Methodologies utilized in literature review papers in the
related fields have been studied. For simplicity, the
methodological approach followed in this paper is adapted
from [29] and [30] where three were three phases
identified, namely, research question statement or focus,
research methodology and research scope. In this paper, six
dimensions or aspects of big data analytics were considered
in the selection of keywords for literature search as follows:
i. The areas of railway infrastructure asset management
in which big data analytics could be applied;
ii. the level of big data analytics in rail network
management;
iii. the trend towards condition-based maintenance;
iv. prognostic health management and smart monitoring
of rail assets;
v. types of big data models, and big data tools and
techniques used for applying these models; and
vi. Blockchain technology for railways and its current
application status.
The selection of these six aspects is based on detailed
discussions among the authors and the determined scope of
this work. The research scope is a literature review on
applications of big data analytics in the railway industry
from the published research articles and reports published
between 1992 to 2019. The reason to choose this time
period is due to that the research area is relatively recent.
To evaluate the available literature, multiple Scopus
searches were carried out initially based on the following
query format. As an initial phase, keywords used included:
(TITLE-ABS-KEY(“asset management” OR “condition
monitoring” OR “data quality” OR “missing data” OR
“machine learning” OR “big data”) AND TITLE-ABS-
KEY(decision*) AND TITLE-ABS-KEY (rail*)). Then the
literature search was progressively adapted and driven by
considerations of the industry issues. Research was
undertaken to further identify the relevant literature and
discussions to focus on the issues, challenges, and
opportunities being faced by industry including potential
application domains. Some of the key challenges identified
within rail asset management included both the trend for
increasing volume and concern about the veracity of data
available for analysis within an asset management system.
Other challenges identified included the need for selection
of appropriate big data analysis tools to match the data
being collected and the trend towards condition-based
maintenance and prognostic health monitoring systems
within rail infrastructure organizations. The volume of data
being collected for import and analysis within asset
management has exponentially increased with the
increasing number of condition monitoring systems
implemented onto the railway network. However, the
veracity of the data is possibly more critical in cases where
rail safety models are required as part of the analysis. Data
quality is inherently impacted by uncertainty and
unreliability of data sources to some extent [31]. As
traditional corporate IT networks may not be compatible
with the introduction of big data analytics, these aspects
were considered to be a part of the focus of the paper and
included in the methodological framework for this paper.
The trend towards condition-based maintenance has also
focused attention on the requirement for intelligent asset
management systems with machine learning capabilities
built in. Additional search terms were identified during this
process for the methodology. As other industries have also
faced challenges in big data analytics, the search terms
were then widened to include other industry sectors such as
the health and manufacturing industries to include case
studies for analysis. For evaluation of available literature on
big data technologies and big data analytics, search engines
through Google Scholar, Web of Science, and IEEE Xplore
were utilized. Based on the article search process as
described above, over 2000 papers were identified for
review and analysis. The selection process was adapted
from Ngai et al. [29] and involved a three-stage approach
where papers were judged primarily on their title, abstracts
and keywords. Two co-authors then debated the extent to
which the paper addressed the six aspects of literature
search criteria identified above. The selected articles were
then grouped in different categories as outlined in Figure 1
based on their primary contributions to each of the
category. It is understood that one paper may cover several
categories, but it is only allocated to the category to which
it contributed the most and counted only once to avoid any

This work is licensed under a Creative Commons Attribution 4.0 License. For more information, see https://creativecommons.org/licenses/by/4.0/.
This article has been accepted for publication in a future issue of this journal, but has not been fully edited. Content may change prior to final publication. Citation information: DOI
10.1109/ACCESS.2020.2967436, IEEE Access
4
VOLUME 8, 2020
duplicated counting. As condition-based maintenance for
railway transportation systems with application of big data
was discussed in [16] and it was identified that the fault
diagnostics and prognostic health management (PHM) for
railway systems is well suitable for a separate paper, the
detailed discussion on these two aspects was therefore not
included in this paper due to the paper length limit
requirement. Finally, 200 articles (including 9 published
through online sources) were cited in this paper to support
discussion and analysis. Some others were excluded as not
being relevant to the focus of this paper or similar contents
have been covered by the ones cited in this paper.
Instead of providing a taxonomy analysis of the
selected articles, the analysis and discussions in the
following sections are more focused on the technical
perspective.
1. Blockchain (Other) (45); 2. Decision Support (41); 3. Missing
Data (42); 4. Condition Monitoring (51); 5. Big Data Analytics
(110); 6. Big Data (75); 7. Asset Management (124); 8. Rail
Infrastructure (169); 9. Infrastructure Maintenance (243); 10. PHM
(Structural Health Monitoring) (40); 11. Asset Management (Oil &
Gas, Utilities) (40); 12. Fault diagnostics and prediction (120);
13. Blockchain (Rail) (14).
FIGURE 1. Grouping of the searched articles in different categories.
III. UNDERSTANDING OF BIG DATA ANALYTICS FOR
RAILWAY ASSET MANAGEMENT
A. BIG DATA IN RAILWAY NETWORK MANAGEMENT
Big data is often referred to as the data as having five
characteristics, i.e. large volume (Volume), fast processing
speed (Velocity), multiple domains (Variety), low density
of the value distributed (Value) and complexity of data
formats (Veracity) [32]. Veracity was initially a term
developed by IBM in 2012 [31] to convey the idea of
uncertainty and unreliability of the data. However, some
researchers would like to highlight the 4 V’s as ‘Value’ is
the results extracted from the data through analysis and
modeling. The value of big data in the decision sciences has
been highlighted by Wang et al. [33] where the different
stages of data capture, curation, analysis, visualization, and
decision-making were discussed. The availability of big
data is regarded as one of the enablers for intelligent
decision making with complex systems [34]. Big data
analysis technologies are required to process and analyze
the data having these five characteristics. In addition, big
data has facilitated the use of techniques such as machine
learning and expert systems for knowledge discovery [35].
Examples of application of these techniques to big data
analysis in other industries include the use of statistical
process control with big data analytics in smart
manufacturing processes [36]. A generic system for
machine learning using big data has been explored by [37].
Another case of using big data analytics has been discussed
by [38] where the existing systems using business
intelligence and data warehouse are limited to handling and
relating unstructured data to structured data. Structured
data, once uploaded from source and analyzed, can be used
as variables in a statistical/machine learning model.
Unstructured data, however, requires further analysis and
decomposition into a set of structured data elements [39].
The significance of big data analytics for asset management
has been described in [40]. In this particular case, the whole
of the asset management system can be supported by the
use of big data analytics and the Internet of Things (IoT).
Volume
Big Data in
Railway
Complexity of
data and various
formats of data
Various data
sources including
trackside & train
data
Value
Velocity
Veracity
Variety
Huge amount of
data per train in
short time
Asset condition;
Asset, equip ID
High speed
processing
FIGURE 2. F ive “V’s”.
Figure 2 provides an overview of some of the
characteristics of big data pertaining to railway network
management. Big data dimensions have been proposed in
different forms to meet challenges in implementation for
various disciplines [41]. Numerous people have increased
the number of “V’s”; however, we focus our discussion on
the five V’s in Figure 2. For railway asset management, the
big data problem can be described in terms of a)
aggregating multiple databases which are individually
manageable and come from different sources, b) individual
datasets that by themselves are too large to be processed by
standard algorithms on legacy hardware [42] and c)

This work is licensed under a Creative Commons Attribution 4.0 License. For more information, see https://creativecommons.org/licenses/by/4.0/.
This article has been accepted for publication in a future issue of this journal, but has not been fully edited. Content may change prior to final publication. Citation information: DOI
10.1109/ACCESS.2020.2967436, IEEE Access
5
VOLUME 8, 2020
relational databases have been designed to deal mainly with
structured data, and little support is provided to semi-
structured or unstructured data [43]. While linking the asset
and equipment identification (ID) with asset condition
information is seen as a key aspect for adding value to
railway asset management systems, this information may be
from different sources and requires aggregation and
processing to adding value. As data is aggregated from
multiple sources, the data may sometimes exhibit heavy tail
behavior and non-trivial tail behavior [44]. There may also
be an imbalance in the datasets where the most important
relationships to be discovered in the datasets are presented
by a small number of examples in the dataset [45]. In a
similar manner to trackside systems, rolling stock on-board
systems may perform the analysis or processing prior to
transmission to trackside-based storage, in a similar
approach to Internet of the Things (IoT) applications. The
challenge here for the asset condition data analysis is to
understand where the raw data analysis is being performed.
B. UNDERSTANDING OF BIG DATA ANALYTICS
Big data analytics is a process of collecting, organizing and
processing large amounts of data to discover useful
information and extract patterns for the purpose of asset
condition prediction, process optimization and decision
making in management to drive better business decisions. A
taxonomy of different processes within big data is provided
by Khan et al. [46]. The focus on big data analysis is to
typically discover the insight into the knowledge that comes
from analyzing the data [47].
Asset condition data on its own without analysis will
not create value for the asset owners who are collecting and
own the data. Once the data is stored and analyzed, it can
create tremendous value [48]. The data analysis or process
can consist of a number of technologies and approaches
such as in-memory analytics, in-database analytics, and
appliances to examine large and varied data sets [49]. There
are six analytical techniques [50] which can be grouped into
four different kinds of analytics as follows [32]:
a) descriptive analytics, which includes reporting/online
analytical processing (OLAP), dashboards/scorecards,
and data visualization. These applications have been
widely used for some time, and are the core
applications of traditional asset management systems;
b) diagnostic analytics, which is used for discovery or to
determine why something happened;
c) predictive analytics, which can suggest what will
occur in the future. The methods and algorithms for
predictive analytics include regression analysis,
machine learning, and neural networks that have
existed for some time; and
d) prescriptive analytics, which can identify optimal
solutions, often for the allocation of scarce resources.
Prescriptive analytics are seen as the future of big data
[51].
To select one of the four kinds of analytical
approaches, an awareness of the concepts of design data
and organic data is required. Hence, in addition to the
evolution of analytics, the ideas of “design data and organic
data” have also been discussed and developed in parallel
[52].
IV. DATA COLLECTION AND MANAGEMENT IN
RAILWAY ASSET MANAGEMENT
A. UNDERSTANDING OF BIG DATA ANALYTICS
A typical railway infrastructure condition monitoring data
collection is shown in Figure 3.
Each of the monitoring systems in Figure 3 is
collecting structured data where the format of the data is
dependent on the type of system being monitored as well as
unstructured data including log files. For example, the noise
monitoring station may be collecting acoustic audio files.
Storage requirements may also be different for different
data types. Field Technician will collect unstructured data
and log files that need to be linked to the monitoring system
structured data. Linking the work order generation from the
CMMS with the correct asset and accessing log files
provides information for later analysis. Within traditional
CMMS systems, the unstructured data including log files
may be stored separately from the structured data within the
CMMS.
Within typical asset condition monitoring systems,
sensors are located either on trackside or within the railway
vehicle undertaking measurements. Typical asset condition
data may include a mixture of related data from high to low
sample rates [53]. The asset condition data sample rate may
be affected by the requirement for the data collection when
equipment passes near a sensor, e.g., a rolling stock vehicle
passes over a track sensor or by a trigger when measured
data passes over a threshold. In this particular example, the
amount of data sent may increase significantly to allow the
evaluation of the trends by an expert system for the
particular threshold that has been reached [54].
FIGURE 3. Integrated rolling stock and infrastructure field data
collection.

This work is licensed under a Creative Commons Attribution 4.0 License. For more information, see https://creativecommons.org/licenses/by/4.0/.
This article has been accepted for publication in a future issue of this journal, but has not been fully edited. Content may change prior to final publication. Citation information: DOI
10.1109/ACCESS.2020.2967436, IEEE Access
6
VOLUME 8, 2020
Condition monitoring sensor systems also provide data
that must be interpreted to turn into alarms [55] and stored
separately in a database for later verification. This
highlights the development of maintenance recording
features of asset condition data which are the precursors for
asset management systems today. Other key considerations
in the introduction of condition monitoring to railway asset
management are ensuring that the technology being used is
as reliable as the asset being monitored and being prepared
for analysis and storage of large volumes of data. Analysis
of the data may have challenges due to the volume of the
data being captured in real-time [56].
Railways across Australia including ARTC and
Sydney Trains have implemented a number of trackside
condition monitoring systems within their network for
monitoring of infrastructure and train assets including the
wheel/rail interface. The Wheel Impact Load Detector
(WILD) system [55] has been utilized for providing
notification when wheel impacts are detected. Condition
assessment is based on visual inspection and some other
measures, e.g., track machines run over the track and
provide a record of the condition of the rail [57]. Integration
of data from condition monitoring systems with track
machine recording can identify the locations where defects
are present and assist in the decision making process [54].
In recent years, Sydney Trains has implemented pantograph
condition monitoring using laser and computer vision
technology [58]. A range of applications using LIDAR and
video recording technology has also been utilized in rail
corridors to provide track degradation measurement data for
prediction of track deflection.
B. DATA MANAGEMENT REQUIREMENTS
There are a number of key requirements within condition
data management including a) data retention requirements
that may be linked to organizational requirements, b) data
accuracy and quality requirements c) data volume
requirements and d) identification of key data required for
decision making and verification purposes. Each of the
requirements can be traced back to the asset management
data model required to meet the asset management
objectives of the organization. Within the advent of
intelligent railway networks, data management has also
been identified as a key challenge in their implementation.
Condition-based maintenance and the requirement for
management of the increasing volume of data has also
required organizations to focus on the development of new
requirements to meet these needs as railway organizations
migrate away from a preventative maintenance approach
[16]. These requirements may include real-time analysis of
streaming data, identification of owner and labeling of
source data in a real-time environment, sharing of data in
real-time to provide for safety hazard notifications as well
as the storage and identification of models to be applied to
the condition-based maintenance data. The condition-based
maintenance data must be converted into information
including details about the quality of the data collected, any
uncertainties, maintenance or operational options [59]. A
case study was undertaken by the Swedish Railways with
the support of Bombardier systems for the collection of
condition-based maintenance data of railway vehicles [60].
A key requirement is to identify at an early stage what
needs to be measured and how to measure, collect and store
the data. The real-time identification of alarms and trend
prediction is identified as requirements [61]. Further
development is occurring within the American Association
of Railways (AAR) to develop interface standards for
condition-based maintenance [62]. A ten-year initiative has
been set up to provide nationwide monitoring and repair
information of freight cars operating in North America.
Data modeling prior to integration may be seen as key to
identifying relationships between the data for rail
maintenance purposes [63].
While standards including IEEE 1451 [64] have been
proposed for regulating the condition monitoring sensor
interface, the asset condition data may be collected and
stored in a proprietary format [65]. Innovative approaches
to converting the data to an open standards format may be
required to extract useful information. Integration of the
condition-based maintenance data using different condition-
monitoring tools used within railway infrastructure is a
critical requirement to ensure the usefulness of the data
[66]. The requirement for integration of data from different
data sources has been extended to other infrastructure areas
where complex decision-making is required to manage a
hierarchy of assets to support maintenance decisions [67].
C. DATA MANAGEMENT SYSTEM ARCHITECTURES
A data warehouse model incorporating condition-based
maintenance using the open systems framework (OSA) is
proposed for an asset management system [68], in which it
is identified that the use of standard terminologies and tools
can assist in more effective use of asset management
information in data warehousing scenarios, with all of the
data being used. This model consists of seven (7) layers as
shown in Figure 4 which is modified for the railway
environment. With the advent of the internet and IoT, the
presentation layer in some cases has evolved to provide a
web-based user interface to match the internet technologies
or via a dashboard. However, the lowest three layers may
still be based on proprietary or bespoke solutions from the
manufacturer or supplier of the condition monitoring
solution. Migrating the top three layers of the OSA-CBM
model to a big data analytics framework can be achieved if
the business process rules for the asset being monitored is
known and understood. In the UK rail industry, an
ontology-based data management approach has been
proposed by the Rail Safety and Standards Board [69]. An
agent-based approach is suggested to relate the identities
and locations of asset equipment.
To utilize the data in an operational environment,
dashboards have been implemented to display real-time
asset condition data [71] to allow for real-time decision
making on asset operational management using descriptive

This work is licensed under a Creative Commons Attribution 4.0 License. For more information, see https://creativecommons.org/licenses/by/4.0/.
This article has been accepted for publication in a future issue of this journal, but has not been fully edited. Content may change prior to final publication. Citation information: DOI
10.1109/ACCESS.2020.2967436, IEEE Access
7
VOLUME 8, 2020
and diagnostic analytics. Figure 5 provides a technical
architecture overview of data collection, integration,
storage, application and presentation. The dashboard
allows for real-time monitoring of agreed key performance
indicators to allow early indication and notification of
divergence from baseline performance. From those blocks
shown in Figure 5, it identifies how to develop appropriate
customer application programming interfaces (API) and
other technologies including enterprise information
integration (EII), enterprise application integration (EAI) as
well as data extraction, transformation and loading (ETL)
tools to pull data from source systems.
#7 Presentation
#6 Decision support
#5 Prognostics
#4 Health assessment
#3 Condition monitor
Data acquisition
#1 Sensor module
Transducer
#2 Signal processing
W
I
D
E
A
R
E
A
N
E
T
W
O
R
K
Proprietary
Application
From
Supplier
Can be
easily
migrated to
big data
analytics
FIGURE 4. OSA - CBM Architecture [70].
Display
Application
Data
Integration
Data
Dashboard
Monitoring
ODS, In-memory cache
Custom API
Legacy systems
BI Portal
Scorecard
Analysis
Management
Data Warehouse
Datamart
EAI
ETL
Packaged Apps
Report
s
EII
Documents
Web Pages
Manual
Files
Surveys
Text
FIGURE 5. Overview of technical architecture for data integration and
application [71].
The type of bespoke applications or services and the
analytical requirements are shown in Figure 6 which is
modified to suit the rail asset data model. Asset condition
information is outputting from each of the applications and
exported to the asset management database. In terms of big
data analytics, analytics engines are required to be tailored
to be compatible with the data formats and calculations
performed at the source data collection points. For the
MongoDB shown in Figure 6, a document store is used
“that does not have any schema restrictions and supports
multi-attribute lookups on records” [72].
In terms of a migration strategy between the existing
traditional relational database systems storing asset
condition data and the NoSQL storage provided by
MongoDB, a hybrid approach can be utilized to minimize
the risk of losing data during the migration [73].
In each of the rail applications identified in Figure 6,
there are separate calculations, in parallel, on the data with
common access to the asset register. Asset condition
information is output from each of the applications and
exported to the asset management database. In terms of big
data analytics, analytics engines are required to be tailored
to be compatible with the data formats and calculations
performed at the source data collection points. Data records
from train describer systems are a valuable source of
information for analyzing railway operations performance
and assessing railway timetable quality [74]. The asset
condition database in Figure 6 may be installed by the
condition monitoring applications provider. The database
choice may be subject to Corporate IT policies within an
organization and utilize databases such as Microsoft SQL,
Oracle and other commercially available or open-source
databases. To ensure compatibility across data sources, a
common data management framework [75] has been
adopted by different railway infrastructure organizations
including CrossRail [76] and Deutsche Bahn [77] to meet
asset master data register requirements [78]. This practice
has been adopted by rail infrastructure organizations within
Australia [79], [80].
FIGURE 6. Data integration platform [81].
Figure 7 provides an idea of the integration of various
data sources within the asset management system. Data
analytics tools using linear, nonlinear and simulation
models such as FlexSim can provide predictive trends for
maintenance management. Data integration is another
critical task that needs to be handled well in asset
management. It requires a common platform to process and
manage diverse sources’ data. The data can be divided at
component level, system level, and operation level data. It
then requires efficient database management using
techniques such as data warehouse for data store and data
mining for classification, etc.

This work is licensed under a Creative Commons Attribution 4.0 License. For more information, see https://creativecommons.org/licenses/by/4.0/.
This article has been accepted for publication in a future issue of this journal, but has not been fully edited. Content may change prior to final publication. Citation information: DOI
10.1109/ACCESS.2020.2967436, IEEE Access
8
VOLUME 8, 2020
FIGURE 7. Example of data source integration for asset management
systems.
D. CHALLENGES OF DATA VOLUME MANAGEMENT
Collection and classification of condition monitoring data
have become an issue in terms of management of large
amounts of data where duplication of resources and data
can occur within the asset management discipline [82].
These challenges and complexities include the volume of
data in collection, tracking of changes to the assets, tracking
and/or recording of maintenance performed on these assets,
and willingness to share data within the organization [83].
For rail assets that may consist of 100’s of kilometers of rail
track including ballast and undertrack infrastructure, the
necessity of spatial information to identify exact locations
in terms of asset condition is an immense challenge.
Converting asset location information to GPS coordinates
may utilize time and resources away from maintenance
tasks. This challenge is not unique to railway organizations
and synergies can be gained from reviewing how other
organizations with spatial datasets are meeting the
challenge [84]. For complex projects, delivery of the asset
management information can be a significant challenge
from a big data perspective [85]. The unstructured nature of
design delivery with information capture can provide a
large amount of unstructured data that needs to be included
within an asset management system for configuration
change management. In terms of the volume of data, the
information technology (IT) infrastructure may require re-
engineering to support the volume of data available for
collection with each asset operation. The impact of using
information and communication technology to achieve the
benefits of big data analytics for rail infrastructure assets
have been highlighted by Takikawa [86]. This may mean a
new platform dedicated to big data analytics [87]. An
example of the complexities in providing access to real-
time asset condition information for rolling stock is shown
in Figure 8 provided from NedTrain [88]. The complexities
can include the interrelationship among multiple suppliers
for an asset. For legacy assets such as older rolling stock,
retrofitting of condition monitoring equipment may be
desirable to meet organizational safety and performance
requirements. In these particular cases, where there is no
train management system (TMS) installed, other modules
are required to be implemented such as the train
communication handler (TCH) connected to the main
equipment room (MER) and the secondary equipment room
(SER) modules to pull data from the diagnostic modules
and sensors. The TCH connects to the network layer which
can be either a cellular 3G, 4G network or Wi-Fi to connect
to the trackside layer which includes network operations
control center (NOC) and databases (DB) for storage of the
condition monitoring data. A further key consideration of
the complexities is the utilization of radio frequency
identification device (RFID) technologies to accurately
identify the location of the asset (GPS plus track location
information) with trackside information using real-time
data analytics [89]. In terms of real-time data collection and
continuous analytics, a paradigm shift may be required
from traditional database methods. This could include
“keeping analytics results in small-sized tables” [90]. An
alternative approach may be to utilize the many task
computing paradigm (MTC) for ensemble-based prediction
methods [91]. This can be extended to utilizing artificial
intelligence or other stochastic methods such as Markov
chains for degradation modeling using asset condition data.
Other railway manufacturers including Bombardier and
IBM have developed big data analytics for train condition
monitoring [92]. These approaches have generally been
focused on the collection of data for fault prediction and
diagnostic explanation. Transportation systems in large
cities such as London are integrating sensor data streams
for prediction purposes and transformation [93]. However,
data volume management is still challenging. The Public
Transport Services Division of the Department of Planning,
Transport and Infrastructure in South Australia has utilized
an IBM Maximo system for asset management and is
progressively rolling out asset condition monitoring
systems as part of transportation expansions [94]. The
Swedish Railways have implemented Maintenance 4.0 with
the requirement for support of large databases [40].
Scalable data structures are recommended as a requirement
to cope with the increasing volumes of data being collected
within a railway environment.
FIGURE 8. Condition monitoring data integration for rolling stock
modified from [88].
E. SUMMARY OF DATA MANAGEMENT AND OTHER
ASPECTS
With the volume of data now available, tools and
technologies have been developed to extract useful
information and patterns from datasets, particularly for

This work is licensed under a Creative Commons Attribution 4.0 License. For more information, see https://creativecommons.org/licenses/by/4.0/.
This article has been accepted for publication in a future issue of this journal, but has not been fully edited. Content may change prior to final publication. Citation information: DOI
10.1109/ACCESS.2020.2967436, IEEE Access
9
VOLUME 8, 2020
spatial data sets where high dimensional data is stored [95].
The complexity of spatial data types, spatial relationships,
and spatial autocorrelation has meant that organizations
need to develop Spatial Data Analysis (SDA) to handle this
complexity [96]. Therefore, intelligent data analysis
methods with application of advanced technologies need to
be developed. This is the focus of research in the past about
20 years and a number of techniques for condition
monitoring data interpretation have been developed (e.g.,
[97]). None of the techniques, however, is perfect and each
is only applicable to certain circumstances. To assist in
application of these intelligent data analysis methods,
technologies such as streaming data analysis have been
developed to cater to the large amounts of data [98].
The idea of missing or incomplete data within spatial
data sets undergoing data mining processes has been
discussed in Brown and Kros [99]. The impact of missing
data on the prediction models can be significant and for this
reason, owners of incomplete data sets may be reluctant to
share the data for asset prediction purposes [100].
In summary, the key challenges facing organizations
managing data on railway infrastructure asset condition
include a) handling of different data types based on systems
being monitored; b) presence of structured and unstructured
data; c) identification of missing or incomplete data sets; d)
handling of volume of data collected with asset condition
systems to sift through for information content and e) asset
data model maturity in measuring conflicting key
performance indicators (KPIs). Display of data sets that are
missing or incomplete which are required for KPIs may
require flags to label the status of these data sets for
potential users of the data sets. A further challenge for
management of the asset data are bespoke applications
provided with the legacy condition monitoring systems that
may not be easily migrated to support a big data analytics
framework. However, rail organizations will need to
consider the future-proofing of asset information using data
analytics technologies [101]. Only a small number of
railway organizations have implemented big data analytics
within their organizations due to the challenges in the
management of asset data [16]. However, there is a
“potential for the application of big data techniques to
manage railway safety” [102]. Big data analytics for
railways include the utilization of sensor technology for
data collection and real-time monitoring of track geometry,
waves, joints and sun curves of rail as well as relative
height, position, wear and defects of catenary wires. Big
data analytics combined with artificial intelligence (AI) has
allowed for real-time prediction of alarms and trends within
the Swedish railways. This can allow an organization to
identify what is critical and what the real problems are.
Further discussions on the application of big data are also
extended to other aspects such as selection of suitable
technologies and tools for some specific applications when
facing challenges.
V. BIG DATA ANALYTICS TECHNOLOGIES AND
CHALLENGES
Advanced analytic techniques have become available with
the advent of big data. At data collection stage, the data to be
collected may include unstructured data from heterogeneous
sources managed through big data analytics engines such as
HADOOP combined with a NoSQL database. At data
analytic stage, the techniques utilized may include regression
analysis, machine learning, deep learning, Bayesian inference
and data mining. The technique selection is closely related to
the data characteristics as well as platforms used for data
collection, management and processing.
A. CONSIDERATION OF BIG DATA ANALYTICS
APPLICATIONS FOR RAILWAY INFRASTRUCTURE
Before we start to consider big data analytics applications,
it would be good to have an overall view of data in
integration such as what is shown in Figure 9. The diagram
gives an overall picture of asset data divided into different
types. The right-hand side of the diagram presents a number
of key applications depending upon the types and context of
the data. That means data availability determines the
potential applications. Identification of the context of data
being collected is critical to the usefulness of the data.
Data Integration
e.g HADOOP,
HANA
Design Data
Structured Data
(Relational SQL
Databases)
High Sampling Rate
(Vibration Signal
Analysis)
Prognostics and
Prediction
Diagnostics
Low Sampling Rate
(SCADA)
Multicriteria
Decision Making
Model-based
simulation
Unstructured Data
(Log Files)
Log Files Verification
Textual Reports Dashboard
Organic Data
Structured
(NoSQL Databases)
High Sampling Rate
(Video Images)
Qualitative Data
Analysis
Low Sampling Rate
IOT Device Logs Status
Unstructured
(NoSQL Databases)
Low Sampling Rate
IoT Measurement
Data
Computational
Linguistics
Natural Language
Processing
High Sampling Rate
(Images)
Pattern
Recognition
Object Detection
FIGURE 9. Considerations of asset data types and context.
The considerations on data collection that impact on
the selection of the big data analytics techniques in railway
industry may include: a) which part of the track or rolling
stock is to be monitored, b) which type of sensor is to be
placed and what kind of data (structured, semi-structured
or unstructured ) is expected from the sensor systems, c)
sparsity of the data to be collected and whether the
collected data reached the SQL database on time or not, and
d) how to deal with bad data (or missing information such

This work is licensed under a Creative Commons Attribution 4.0 License. For more information, see https://creativecommons.org/licenses/by/4.0/.
This article has been accepted for publication in a future issue of this journal, but has not been fully edited. Content may change prior to final publication. Citation information: DOI
10.1109/ACCESS.2020.2967436, IEEE Access
10
VOLUME 8, 2020
as missing tag reader identification). Each of the
considerations above can be addressed initially with the
concept of “design data” where the data requirements are
included in the design of the system to be monitored.
However, as the system is monitored in a real-time
environment, the concept of “organic data” is introduced
which can include sensor data, sentiment data and various
types of machine data that can be characterized by context.
The data context is important in providing the links or
relationships between each of the nodes of big data [103].
Big data analytics techniques such as machine learning and
Bayesian inference could be used to read monitoring data
and also learn data context and correlation (e.g., rolling
stock identity vs wheel temperature) for efficient
maintenance planning and decision-making.
There are new questions on data analytics in terms of
accuracy such as: a) which technique accurately predicts the
condition of the asset; b) how to deal with missing data in
the modeling and how it will affect the trends produced
from the modeling and c) how to manage the data flow
where the data may be intermittent based on equipment
availability and timeliness requirements of the data.
Veracity of the data is of key importance within railway
asset management due to the requirement for safety as a
priority where uncertainty and reliability of data sources
can arise. While large data sets are assumed to be better,
they may contain systematic biases or have large amounts
of missing information, and even missing key variables
which can be magnified with the size of the dataset [104].
Big Data sets may be collected under some complex and
unknown measurement process [105]. This process needs to
be included in the process of design data. Alternatively, the
big data set may have been collected for a different purpose
and an analysis is being performed to see if any inference
can be drawn from this data set [39].
In terms of diversity of data, the key challenges are in
matching the diverse data sources to a data management
model. Alternate approaches using heterogeneous data
integration are provided by [106] where multiple kernel
estimators have been proposed to develop an understanding
of the underlying data structure for integration into a data
model. Data sources such as equipment maintenance logs
may not match the structure of the other data sources being
imported. Analytics approaches using sentiment analysis
and natural language processing may be required for
information extraction to associate the unstructured data
with a failure event [107]. Further development is required
for utilization of these techniques within engineering asset
management. Computational analytical solutions,
particularly using unstructured data, may not yet be
sufficiently developed for safety-critical infrastructure
systems [108]. While some failure modes can be identified
during the design phase for a system, failure modes may be
identified after operational trials [109] which mean
additional separate sensors to be used for collecting failure
information that are not part of the asset system model. A
way of achieving data integration with diverse data sources
is to utilize scenario-driven data modeling (SDDM) [110].
This method also allows for the identification of gaps
within a data management model to allow adaptation for
missing data elements. This approach can also allow for
event-driven data collection where the collected data can
change depending on an event [111]. If failure modes can
be identified after the event, the use of event-driven data
analysis and scenario data modeling can help identify gaps
within the existing data management model to identify
patterns in failure data. The required changes to the data
management model can then be modeled using universal
modeling language tools (UML) to assess impacts on the
data collection requirements. Where failure modes are not
readily identifiable, then maintenance staff may need to
collect all of the data and search for patterns within the
data. An example of an ontology method for data
integration with big data is provided by Eine et al. [112]
where the relationships can be mapped using equivalence
rules to automatically deduce the relationships between
different data sources. This method can also be prepared
with the approach of using ontology for data management
discussed earlier.
In the railway asset condition monitoring example, the
trackside data collection site can be designated as the
primary site while a large amount of data collection process
can be designated as the secondary site where the data
integration and mapping begin. The other part of the
integration process is in identifying an authoritative source
uniting overlapping datasets from separate primary data
sources, which may supply redundant or conflicting values
[113]. For each of the databases (DB) in Figure 6, a
separate service is required to pull the data from the DB to
the primary source (Global Model). Each of the separate
services may utilize proprietary technology based on the
origin of the primary source. For example, if the trackside
sensor produces sound files with text data, a ‘Service’ that
is suited to importing these types of files may be utilized for
import facility to the Global Model. The RailBAM™
application, for example, produces wave files with text files
based on the train consistently for each train pass.
Another example of the type of bespoke applications
or services and the analytical requirements is given in
Oweis [21] where, in each of the bespoke applications
identified, there are separate calculations in parallel on the
data, with common access to the asset register.
An illustration of a modeling framework for track
maintenance is shown in Figure 10 [114]. From this figure,
a model is selected for the particular asset being assessed
and the live asset condition data is applied to the model
with an evidence-based track working model as its output
which is stored in the asset management system.
The collection of rail inspection data has increased in
volume and quality in recent years and has been presented
as a big data analytics challenge [115]. Rail inspection
processes have developed from two approaches: a) on-
board sensors mounted on rolling stock and b) track-
mounted sensors. In these approaches, further discussion is

This work is licensed under a Creative Commons Attribution 4.0 License. For more information, see https://creativecommons.org/licenses/by/4.0/.
This article has been accepted for publication in a future issue of this journal, but has not been fully edited. Content may change prior to final publication. Citation information: DOI
10.1109/ACCESS.2020.2967436, IEEE Access
11
VOLUME 8, 2020
required to ensure data format standards and requirements
can be met by the different manufacturers within the rail
industry.
System level
Operation level
Component level
Track n Degradation
Model
……
Track 1 Degradation
Model
Component Maintenance
Model
Component Inspection
Model
Track n Working Model
……
Track 1 Working Model
Track System State
System Inspection
Model
Railway System
Operation
Track Occupied
Inspection
Accident Model
State Update
Component Maintenance
Maintenance
Decision
Decision Model
Track States
Line Closure or Speed
Restriction
Maintenance
Decision
FIGURE 10. Modeling framework for track maintenance modified from
[114].
The considerations outlined above need to be
evaluated for each of the asset classes and types included
within the railway infrastructure asset management model.
B. BIG DATA ANALYTICS TECHNOLOGIES
A short review of big data technologies can be found from
[116][119]. These technologies are related to the
challenges being faced by using big data, namely, volume,
complexity and high-speed processing. Advanced analytics
technologies are developed to support predictions using
scenario developments rather than probabilities as in more
traditional analytics technologies. Analytics is about the
discovery of new or existing relationships within the data
sets and making sense of the data for modeling purposes.
Multiple analytics technologies may be utilized within big
data applications to support the variety of heterogeneous
data being collected. In-memory analysis can be performed
during the data collection stage to determine the value and
relationships of the data prior to storage [48]. These
technologies can be grouped into two key categories of a)
storage and b) querying/analysis [120]. Storage relates to
the “persistently storing and managing of large-scale
datasets” [121]. The technologies associated with data
storage and management are also closely related to the
frameworks and platforms utilized, e.g., HADOOP
frameworks, HDFS and non-relational databases such as
NoSQL. Due to the volume and speed, incoming data may
be initially split and stored across several machines. Big
data storage technologies may be required to split the
incoming data across several machines simultaneously for
storage and analysis. These technologies and tools are
discussed further in Section V.D. Querying and analysis
relate to the use of analytical technologies or tools to
inspect, transform, and model data to extract value
(information). At this stage, big data analytics may be
divided as descriptive, diagnostic, predictive and
prescriptive analytics with the purpose of application from
information learning to insight and further to foresight
(optimization) of the process. Technologies utilized for the
querying and analysis include data mining, clustering,
knowledge discovery, machine learning, MapReduce,
Massively Parallel Processing (MPP), Multi-Dimensional
On-Line Analytical Processing (MOLAP), visual analytics
and statistical models [122][125]. Table I provides an
exemplar of big data analytics technologies for reference.
In each of the examples provided in Table I, aspects of
the analytics technologies may be used in more than one
example. Machine learning has been utilized across
multiple analytics technologies to increase efficiency. A
key challenge being faced by railway organizations in
introducing asset condition monitoring sensors is issues
with scalability, security, economics, and engineering.
Currently, connectivity between railway condition
monitoring sensors utilizes a number of mediums for
connectivity (see Figure 8). This means that the IoT
structure can be regarded as a centralized structure. This can
impede the flow of communications between sensor devices
and the analysis processing units. To address the limitations
of a centralized structure, Blockchain technology has been
proposed to assist in decentralized computation and asset
condition monitoring by providing a distributed ledger that
can record transactions between two parties efficiently and
in an authentic way [126]. Further discussions on
Blockchain and its applications to the railway distributed
processing are provided below.
TABLE I
EXAMPLES OF BIG DATA ANALYTICS TECHNOLOGIES
Big Data
Analytics
Technologies
Examples
Descriptive
Technologies
Examples include OLAP, EAI, EII,
ETL with dashboards for display and
real-time reports. Data visualization
allows for condensing big data into
smaller, more useful nuggets of
information. Mining historical data
to look for the reasons behind past
success or failure.
Diagnostic
Technologies
Variety of technologies such as drill-
down, clustering, data discovery,
data mining and correlations. It is to
examine data or content to answer
the question “Why did it happen?”.
Predictive
Technologies
Variety of statistical, modeling, data
mining, machine learning, and
sentiment analysis technologies
(natural language processing, text
analysis, computational linguistics)
to study recent and historical data.
Predictive analytics can be
probabilistic in nature.

This work is licensed under a Creative Commons Attribution 4.0 License. For more information, see https://creativecommons.org/licenses/by/4.0/.
This article has been accepted for publication in a future issue of this journal, but has not been fully edited. Content may change prior to final publication. Citation information: DOI
10.1109/ACCESS.2020.2967436, IEEE Access
12
VOLUME 8, 2020
Prescriptive
Technologies
Variety of technologies such as
graph analysis, simulation, complex
event processing, neural networks,
recommendation engines, heuristics,
and machine learning. Prescriptive
technologies can be considered to be
predictive technologies with
additional components for actionable
data and feedback that tracks the
outcome produced by the action
taken to provide optimization and
then suggest decision options to take
advantage of the predictions.
C. BLOCKCHAIN TECHNOLOGIES
C.I. CURRENT APPLICATION STATUS OF BLOCKCHAIN
TECHNOLOGY IN RAILWAY INDUSTRY
Blockchain can be described as a shared distributed ledger
[127] with encryption to provide authenticity [128]. For
distributed parallel processing of data coming from multiple
sensors, there is a problem arisen that is the trust between
sensors and processing components and the generation of
digital signatures for each of the sensor devices. Blockchain
can be utilized to build a trust relationship using a smart
contract concept [129]. An alternative decentralized system
to register and assign IoT devices to an owner based on the
Blockchain technology has been proposed by Ghuli et al.
[130]. In this approach, initial ownership is provided by the
manufacturer of the device and then being transferred to an
owner based on Blockchain technology. However, it is not
clear with this approach if replacement of the failed devices
would have occurred with updated registration of devices.
Blockchain can be utilized to provide parallel
processing and communications architecture where the data
flow is decentralized and security is improved [131]. This
can also decrease response times where data flows can be
directed to the closest node where the data may be utilized.
Data on track status such as faults or obstacles including
location information can be shared between the track
devices and the train (rolling stock) without first sending the
information to a central device (Rail Control Centre). The
same data can be sent to the central device in parallel using
Blockchain technology to verify the receipt of the data
[132]. The types of applications currently being reviewed by
Deutsche Bahn with Blockchain include safety applications
where track obstacle reporting is provided directly to trains,
and track condition monitoring and maintenance systems
[133]. Enterprise asset management systems such as the
IBM Maximo product currently support Blockchain for
asset management applications [134]. Pacific National
within the Australian rail freight network has been utilizing
Blockchain for supply chain management of perishable
goods [135]. This involves the tracking of the perishable
goods across the network with multiple partners. The
approach adopted within the rail industry has utilized a
similar approach to that outlined for the energy sector in
[136]. Other examples representing the current application
status of Blockchain technology in railways are summarized
as shown in Table II.
TABLE II
SUMMARY OF CURRENT APPLICATION EXAMPLES OF BLOCKCHAIN IN
RAILWAY INDUSTRY
Name of
Company or
Organization
Examples
Swiss Federal
Railways
Tested a worker identity management
system ( [137], [138]).
Blockchain in
Transport
Alliance
(BiTA)
Making efforts to develop new
framework and standards for
transportation companies ([139],
[140]).
Go-Ahead
Group Plc in
the UK
The company is reportedly partnering
with Blockchain start-up DOVU to
launch a tokenized, Blockchain-based
rewards system for its customers
([141]).
Russian
Railway (RZD)
Launched the Freight Transport, an
electronic trading platform
underpinned by Emercoin and more
than 5000 freight consignments
ordered via the platform over a nine-
month period [142]. The company is
going to implement Blockchain
applications for ticket sales and smart
contracts all in crypto ( [143], [139]).
Bourque
Logistics in the
US
Unveiled the RAILChain™
platform to explore the application of
Blockchain technology for its rail
shipper clients. The platform is
designed to enable shippers to
securely exchange bill of lading
information, as well as settle freight,
repair and lease costs using smart
contract technology [144].
The State
Railway of
Thailand
Is investing Blockchain to manage
signalling, passenger information
systems, ticketing and goods delivery
[139].
Shenzhen
Metro in China
Launched the first-ever Blockchain-
based digital invoicing system in
March 2019 [145].
C.II. CONSIDERATION OF THE FUTURE APPLICATIONS
OF BLOCKCHAIN TECHNOLOGY IN RAILWAYS
The data flows in a railway network can be expressed using
an IoT architecture as shown in Figure 11, in which the Rail
IoT Connections represent parallel distributed connections
between track, trains and station staff; and the Rail Network
Connections represent the centralized communications
path. Thus, the asset condition data collected from trackside
sensors may be received twice and arbitration may be
required to distinguish between the two paths for the same
data. However, the data available in a single public
distributed ledger shared among several parties can be more

This work is licensed under a Creative Commons Attribution 4.0 License. For more information, see https://creativecommons.org/licenses/by/4.0/.
This article has been accepted for publication in a future issue of this journal, but has not been fully edited. Content may change prior to final publication. Citation information: DOI
10.1109/ACCESS.2020.2967436, IEEE Access
13
VOLUME 8, 2020
reliable than multiple centralized databases. A time-
stamped version of the ‘data’ can be available on the
distributed ledger to show the arrival of the data at various
devices for audit purposes.
The data flows can be enabled more effectively with
Blockchain technology. For example, the verification of the
asset condition status by the asset management system after
maintenance intervention by trackside staff can be achieved
using Blockchain technology and IoT network in Figure 11.
Each of the devices can be arranged into a peer-to-peer
(P2P) connection where the Blockchain is a decentralized,
distributed ledger (public or private) of different kinds of
transactions arranged into each device. The data cannot be
altered without the consensus of the whole network. Track
location information with vehicle RFID information can be
shared between trackside staff and the Rail Operations
Centre for decision-making with the assurance that the
information has been validated without manual verification
processes being utilized. Work orders can be sent directly to
maintenance staff for allocation of maintenance resources
without human intervention. The utilization of Blockchain
technology allows the verification of the source of the
information as a trusted source in real-time with distributed
processing and storage of the data by big data analytics.
Different communication mediums such as WIFI, 4G
mobile and trackside communication systems can be utilized
directly as available without the concern about the security
of the communication mediums.
FIGURE 11. Rail IoT Architecture [146].
Given the description above, it can be foreseen that the
applications of Blockchain technology in railway network
management will be promising in the near future. These
include smart maintenance with better data tracking for
improving safety and reducing service costs, transit
payment and ticketing for improving service and cutting
customer costs. To sum up, Blockchain technology is at the
start-up stage [147]. “The opportunities are exciting but
companies will need time to familiarize themselves with the
technology and to identify the best prospective application
areas which are supported by a universally-accepted set of
standards” [142]. These, however, are still in the
developing process.
Blockchain has its strengths and advantages in
application but the speed and scalability would be
remaining as a big concern in developing Blockchains for a
busy network.
D. BIG DATA ANALYTICS TOOLS AND OTHER
ASPECTS
Tools have been developed to manage the five
characteristics of big data described above and the five-
stage process which forms two main sub-processes of data
management and analytics [148]. These tools can allow for
parallel processing and loading of chunks of data to assist
in key activities for the handling of the large volume and
diverse types of data. Examples of these tools include
MapReduce which allows, for example, splitting of the
input data-set into independent chunks which can then be
processed in a completely parallel manner [69], [107],
[149]. Python is commonly used when data analysis tasks
need to be integrated with web apps or if statistics code
needs to be incorporated into a production database [150].
The utilization of Python for big data analytics is generally
via a framework using Python libraries and tools [151].
Each of these Python frameworks has characteristics that
are designed for specific models. Other Tools such as
Regular Expressions (RegEx) can be used to detect and
repair errors for data input as part of the pre-processing
stage [152]. The requirement for expert systems or artificial
intelligence was identified to pre-process the large amounts
of data [70]. This approach has been further developed with
the concept of machine learning [153]. The Machine
Learning Library (MLIB) is utilized to provide a number of
machine learning algorithms that can be implemented
within big data infrastructure [154]. However, other
machine learning tools such as backpropagation can be
parallelized with big data analytics such as MapReduce
[155]. In this example, the original data information is
maintained in the data subset, which can be useful for
verification purposes. For the types of data collected by
condition monitoring systems, different tools may be
selected to process the data from the raw to the processed
data stage (e.g., semantic analysis for the raw text files).
Semantic analysis for message analysis can be done with
big data analytics [156] and semantic analysis can be
performed using support vector machines (SVM) where the
accuracy of sentiment classifications can be improved
[157]. The semantic analysis process developed could be
applied to unstructured data such as log files or status
messages provided as part of the asset condition monitoring
process.
E. BIG DATA ANALYTICS TOOL SELECTION AND
CHALLENGES
The challenges for the use of big data within an asset
management environment relate to a) storage of the diverse

This work is licensed under a Creative Commons Attribution 4.0 License. For more information, see https://creativecommons.org/licenses/by/4.0/.
This article has been accepted for publication in a future issue of this journal, but has not been fully edited. Content may change prior to final publication. Citation information: DOI
10.1109/ACCESS.2020.2967436, IEEE Access
14
VOLUME 8, 2020
and also heterogeneous types of data; b) management of the
data collection process to ensure the accurate and reliable
collection of data; c) managing the changing relationships
between the data and the assets that the data describes and
d) inconsistency and incompleteness of the data [158].
These challenges include, for example, handling of
heterogeneity, dealing with inconsistency and
incompleteness, merging data, timely process and analysis,
and ensuring privacy and data ownership [159]. Track
geometry measurements (see Figure 10) in particular are
described as requiring a high sample rate to capture the
trends for prediction and analysis purposes. Analyzing large
volumes of data in real-time may require the combination
of in-memory processing with analysis by advanced
machine learning techniques. Companies have had to deal
with profound changes in the use of technologies for big
data analytics [160]. The selection of different
unconventional tools such as Python and MongoDB to meet
big data challenges has meant that companies have had to
adapt to changes in the enterprise data model and data
analytics [161]. The use of JavaScript Object Notation
(JSON) data format with NoSQL databases such as
MongoDB allows for efficient analytics functionality [162].
The selection of tools for managing big data is often
described as a collection of related techniques and tool
types. Those are usually utilized for predictive analytics,
data mining, statistical analysis, and database management
that support analytics [32] such as MapReduce with in-
database analytics, in-memory databases, and columnar
data stores. As most traditional tools and algorithms are
regarded as inefficient, the development of new algorithms
is required. The design of an asset management system at
this stage may appear to be different from that expected for
a conventional asset management system using off the shelf
applications and tools. Python and R programming
languages can be used to implement machine learning (ML)
techniques to support complex degradation models
requiring sufficient data. This can help improve the
outcomes of complex maintenance decisions [163]. The
benefits of migrating to new tools, particularly in a data-
rich environment can be large [164].
While big data has been applied to transit forecasting
and ridership forecasting within rail systems [27], the use of
big data analytics for predictive maintenance is in its
infancy. Network Rail has utilized Deloitte’s suite of cloud-
based analytics tools to provide real-time timetable
information for timetable management [165]. The
development of advanced analytics for train delay
prediction using exogenous data is provided by Oneto et al.
[166] where multivariate statistical concepts are
implemented using big data analytics tools, and
improvements in train delay prediction using advanced data
analytics including multivariate statistics over traditional
delay prediction methods are discussed. Further work is
suggested for inclusion of railway asset condition data to
improve prediction accuracy. Within rail infrastructure
environments, energy efficiency has been given increasing
priority to improve sustainability. An example of predictive
analytics using artificial intelligence (AI) for rolling stock
power consumption on the railway network has been given
by Furutani et al. [167]. While AI has been regarded as
being in its infancy within rail infrastructure networks,
impetus is being provided to roll out AI to the rail sector.
The key challenge identified here is the real-time
processing of the stream of data from sensors. This is also
the case for other industries such as health care and
infrastructure organizations [10]. A key item is identified as
the maturity of the available big data platforms and
analytics software which are still regarded as being in their
infancy [125]. This includes concerns about scalability to
meet increasing data volumes and substantial real-time
requirements for asset condition monitoring. The use of big
data analytics for transit forecasting analysis of human
mobility has been given by [168]. This has applications for
transportation passenger utilization prediction using
passenger mobile phone location data. Trials of this
approach are currently being conducted in the US and
Europe to assist in passenger load forecasting and
congestion reduction. Access to passenger information can
be difficult when evaluating the value of rail infrastructure
investments. There is currently a digital transformation
focus on railways in terms of utilizing big data for better
management of resources including passenger utilization of
assets [169]. The recent development of European standard
(EN12896) for a reference data model for public transport
information has provided for sharing of timetabling, fares,
operational management, real-time data, and journey
planning across railways [170]. In addition, the AAR has
published a set of standards for data handling to meet
railway operational requirements [171].
In terms of rail maintenance applications, an example
of using machine to machine (M2M) procedures with a
customized condition-based maintenance platform to
handle big data challenges has been proposed by Palem
[172]. For this application, real-time data collection and
analysis have been emphasized for fleet maintenance
prediction purposes. A reference architecture (platform) is
proposed where Python and R have been proposed for the
data analytics engines. In particular, R-frameworks can be
utilized where the analyst does not need to know the
underlying intricacies of the big data architecture or data
model [173]. A decision support system using smart data
for a railway metro system is provided by He et al. [174].
In this case, the use of data from different sources for the
decision support system is described and compared with the
innovation requirements to support a large-scale railway
metro system. This has supported a change in focus from
equipment-centric to asset-operation centered. The concept
of smart maintenance decision support systems is being
trialed in the US where the data analytics is used to extend
the linkage between the analysis of condition monitoring
data and statistical trending with prediction and simulation-
based scenarios [175]. The availability of large volumes of
data to build accurate simulations of complex systems can

This work is licensed under a Creative Commons Attribution 4.0 License. For more information, see https://creativecommons.org/licenses/by/4.0/.
This article has been accepted for publication in a future issue of this journal, but has not been fully edited. Content may change prior to final publication. Citation information: DOI
10.1109/ACCESS.2020.2967436, IEEE Access
15
VOLUME 8, 2020
extend the predictive maintenance concept within large
infrastructure organizations.
Challenges have also been identified in capturing
changes in assets particularly when delivering a major piece
of infrastructure such as a new railway in the era of big data
[85]. Related IoT technologies are being deployed within
industrial asset management environments to provide the
connectivity for big data [176] through the improvement of
the data acquisition process for condition monitoring
sensors. The choice of big data analytics tool is then based
on the IT & technology policies of the particular
organization undertaking the data analytics development. In
addition, the security and privacy of the stored data have
become a challenge as the volume of data has increased.
Automation and centralisation of the processing of the data
with the added security and privacy requirements may
require a re-think on IT strategy [177]. Where datasets are
heavily skewed, changes in the clustering algorithms such
as DBSCAN for parallel processing may be required [178].
In addition, a reference architecture may need to be
developed for implementation [179]. There must be very
clear requirements for implementation of big data analytics
[180]. Identifying the underlying structure and key
information is important in achieving the value of big data
analytics.
In summary, rail organizations are focusing on the
development of big data analytics framework for railways
while technology choices are being made by individual
organizations based on requirements [181].
VI. OVERVIEW OF BIG DATA ANALYTICS APPLICATION
TO RAILWAY INFRASTRUCTURE
Big data analytics within the railway environment is still in
its infancy as highlighted earlier. Key applications of big
data analytics within railway infrastructure include decision
making for infrastructure maintenance based on available
condition monitoring data as well as operational
performance and train timetabling data. The data processing
requirements for the collection of decision-making data
include integration of data from heterogeneous sources
where low latency requirements for decision making need
to be provided through parallel architectures. These include
track circuit status monitoring where data flows may be
triggered on train movements through track sections in
short time periods of seconds or more. Other approaches
include the collection of rolling stock condition information
for condition analysis to provide a plan for maintenance
when the rolling stock reaches a destination.
A. BRIEF DISCUSSION ON APPLICATIONS OF BIG
DATA ANALYTICS FOR PROGNOSTICS (PHM)
As organizations move towards predictive maintenance
programs away from planned and reactive maintenance, the
importance of prognostics and health management (PHM)
has increased. Detection of anomaly patterns within PHM
can be used to detect the existence of a fault before a failure
happens [182]. Key aspects of the prognostics models
available have been identified by Peng et al. [183]. Hybrid
approaches using machine learning using a combination of
two or more algorithms to model the system were
suggested. The four dimensions of prognostics, i.e., a)
sensing, b) prognosis, c) diagnosis and d) management have
been described by Kwon et al. [184] where prognosis
requires additional data including maintenance history,
operational and performance parameters that were not
previously available with sensors or diagnosis but are
available with data integration from heterogeneous sources.
PHM has been described as a recent advance that can allow
for a better understanding of the degradation process of a
component and a system so as to extend and manage
efficiently the life duration of industrial systems [185]. The
PHM methodology allows for the utilization of remote
sensing data, condition monitoring data and, the
interpretation of environmental, operational and
performance parameters to indicate systems’ health status
[186]. The availability of data from multiple sources such
as condition monitoring systems and reliability analysis
systems has allowed for the use of regression, degradation
model, support vector machines (SVM), neural networks,
and other Artificial Intelligence (AI) techniques particularly
with decision support and decision-making models for
PHM [164], [187]. A discussion of the model-driven and
data-driven approaches for PHM using artificial
intelligence has been conducted by Schwabacher and
Goebel [188], where verification and validation of the
diagnostic models are difficult to achieve before
deployment. A key reason for this is that the failure modes
can occur which are not identified in the model-driven
approaches prior to deployment as well as the lack of
historical state (condition) data that can impact on the data-
driven approaches. Choice of algorithms to be used for
PHM will be based on the level of complexity required of
the model and the amount of noise presented in the data
[189], [190]. Another key challenge faced in using PHM
methods is in “taking uncertainty into account” [191].
However, a key challenge with machine learning in
prognostics health monitoring based on analysis of asset
condition data is the handling of high dimensional data sets.
Using domain knowledge that faults and their implications
correlate to a type of information contained in an asset’s
life cycle data and are translatable to a type of domain
knowledge representation with an entropy measure can be
utilized to provide a dimension reduction framework [192].
Alternative approaches for dimension reduction using
principal components with machine learning have been
identified by Gorban and Zinovyev [193]. Table III below
provides a summary of the application of big data analytics
with prognostics health monitoring.
TABLE III
SUMMARY OF APPLICATION OF BIG DATA ANALYTICS WIT H PROGNOSTICS
HEALTH MONITORING
Application
Type of
condition
monitoring
sensors
Analytics approach

This work is licensed under a Creative Commons Attribution 4.0 License. For more information, see https://creativecommons.org/licenses/by/4.0/.
This article has been accepted for publication in a future issue of this journal, but has not been fully edited. Content may change prior to final publication. Citation information: DOI
10.1109/ACCESS.2020.2967436, IEEE Access
16
VOLUME 8, 2020
Rail condition
monitoring
Vehicle-
mounted sensors
Real-time analytics
comparing with a
model for
diagnostics of
deviations from
track geometries.
Environmental data
can be added to
predict when track
geometries may
require attention
Train
performance
monitoring
Trackside and
vehicle-mounted
sensors; wheel
flats measured
by impact
sensors mounted
on track
Real-time analytics
comparing with
different models for
threshold alarms.
Hybrid data
approaches used to
enhance existing
models. Alarm
threshold
information can be
provided to train
operators in real-
time for planning
maintenance
intervention
Traction
Overhead Wire
Sag measurement
Trackside and
vehicle-mounted
sensors; sag
measured by
deviation from
known height
using CCTV
images
Real-time analytics
of sensor and
CCTV images to
measure traction
wire sag while train
is passing under
overhead wire.
Environmental data
can be added to
predict when
overhead wire sag
could reach a
threshold.
Railway Points
monitoring
Trackside audio
measurement
sensors
collecting audio
data to
efficiently detect
and diagnose
faults in railway
condition
monitoring
systems.
Real-time analytics
of sensor and audio
data to detect
anomalies while a
train is crossing a
set of points using
support vector
machines (SVMs).
Foreign object
detection on
track
Vehicle-
mounted sensors
using CCTV
images to detect
objects in front
Real-time analytics
using convolutional
neural network
(CNN) of CCTV
images to detect
on track
objects while train
is travelling on
track Radar and
location data can be
added to predict
when objects may
be present.
B. DATA MINING
Traditional techniques such as data mining have also been
utilized with big data analytics to provide improved
outcomes for condition-based maintenance and rail
infrastructure asset management. Predictive analytics using
data mining with big data has been identified as a future
need where real-time analysis will be the key to meet
scenario-based analysis techniques [194]. The utilization of
technology for onboard train equipment can assist in the
collection of condition data for analysis using data mining
techniques. In the example provided by Sammouri et al.
[195], the association of Spatio-temporal data is utilized to
determine if a significant asset condition related event has
occurred to trigger the mining of the data associated with
this event for analysis. Condition-based maintenance may
require a continual monitoring of an asset leading to
volumes of data requiring real-time analysis. A challenge
for rail operators is in being able to predict a defect
beforehand so that the failure does not occur in operation
and does not cause a train delay [196]. To meet this
challenge, a spatio-temporal-nodal model has been
proposed to analyze the interdependencies of various rail
systems and identify root causes for failures and system
behaviors.
An approach using big data streaming analysis has
been provided by Fumeo et al. [98]. In this approach, the
continual monitoring is akin to a prognostic approach
where triggering of maintenance actions occurs when
degradation is detected. However, older data in some cases
may no longer be available to revise earlier suboptimal
models. Further strategies including data mining may be
required to assist in the data analysis process. An example
provided by Cannarile et al. [197] outlines how data mining
for a large number of assets can be used to optimize the
maintenance strategies for a particular group of railway
infrastructure assets. Population-based strategies were
compared with cluster-based strategies for a group of
30,000 switch point machines. While improvements in
analysis were reported when using cluster-based strategies,
further work in the selection of the starting point or
decision variables for the cluster analysis is required. This
is consistent with the traditional usage of data mining and
clustering algorithms. It may be difficult to choose an
appropriate clustering algorithm for use in specific big data
sets without detailed knowledge of the characteristics of the
big data set [198].

This work is licensed under a Creative Commons Attribution 4.0 License. For more information, see https://creativecommons.org/licenses/by/4.0/.
This article has been accepted for publication in a future issue of this journal, but has not been fully edited. Content may change prior to final publication. Citation information: DOI
10.1109/ACCESS.2020.2967436, IEEE Access
17
VOLUME 8, 2020
C. SUMMARY
Big data analytics with the increased availability of data has
provided tools for the analysis of more complex systems,
while also improving the outcomes for analysis of existing
systems [187]. However, the increased availability of
selection of tools has complicated the task of selection of
the right tools for big data analytics with rail infrastructure
assets. The selection of methods for a practical application
will be based on the availability and quality of the data for
the analytics process. However, there is a tendency towards
the hybrid approach due to the limitations with the model-
driven approach, in which the model-driven approach is
used as a starting point for the hybrid approach. While the
volume of big data is increasing, the challenge is in being
able to select the right data to describe the asset condition
information and also in undertaking degradation modeling
analysis. Selection of machine learning techniques may
impact on the scalability of the applications for increasing
data volumes and high dimensionality. Each selected
algorithm may have a point where performance starts to
drop with increasing volumes of data. Hybrid approaches to
analysis techniques may be required. Uncertainty is built
into the data and needs to be accommodated in the
collection and analysis of the data. Big data analytics
algorithms need to be developed further in order to make
full use of information hidden in big data. Techniques are
crucial to reducing uncertainty in big data analytic process
for improving the performance of prediction. Further
research has been proposed for the utilization of deep
learning techniques to recognize model limitations and
improve the prediction accuracy when data quality is poor,
or missing data patterns are evident within the data
collected where the concept of platform analytics including
machine learning and expert systems is introduced for
optimization of timing and types of maintenance to be
performed for different rail infrastructure assets [86].
VII. CONCLUSIONS
Big data analytics for rail infrastructure management is not
fully mature, while condition monitoring systems are in
general use for asset condition classification purposes.
However, there is no agreed common method of application
or use of algorithms for asset condition assessment. There
is also no agreed interface for integration of the types of
information utilized for measurement of the asset condition
with asset management systems for big data analytics.
Issues, challenges, and future research directions are
summarized below.
A. GAPS AND ISSUES IN CURRENT RAILWAY ASSET
MANAGEMENT WITH APPLICATIONS OF BIG DATA
ANALYTICS
The collection and updating of asset condition data at
organizational levels of control are extremely time-
consuming. Collection and classification of condition
monitoring data is an issue in terms of management of large
amounts of data and duplication of resources within the
asset management discipline. This may involve the
collection of historical data with current component failure
data to uncover new patterns for prognostics applications. A
change in approach from model-driven to data-driven
algorithms may be required to analyze data/information
collected from the system(s) and the operational
environment. The addition of unstructured data into the mix
and the requirement to identify relationships for decision-
making purposes can increase the complexity of the
challenge. Timelines of the analysis may be difficult with
the volume of the asset condition data being collected.
Smart frameworks with machine learning combined with,
for example, fuzzy systems models can be used to optimize
the decision-making based on the data-driven inputs [199].
However, the proprietary nature of the condition
monitoring systems used to collect the data can restrict the
ability of end-users to integrate condition monitoring data
from different sources to provide aggregate views of asset
condition. This has led to the development of open systems
architecture for condition-based maintenance. Condition
monitoring systems data collection can provide timely data
for maintenance planning at scales appropriate to a variety
of maintenance activities. The problems of quality control
and quality monitoring from the perspective of condition
monitoring data classification by non-experts require
development of new approaches. Disparate and large
registers of all asset types, most of which are remote, are all
subject to annual verification primarily motivated by
financial accounting concerns. The volume of data
collected is so huge that it has become humanly impossible
to do any intelligent data analysis. Hence, there is an
increased reliance on automated big data analysis tools
which requires an increased level of testing and
verification. A lot of managerial tasks including research
works need to be done in a large railway organization
before a condition-monitoring database is established and
can be used in assisting in asset management decisions.
Evidence-based asset management models are
currently being developed to meet the requirement for
justification of investment decisions in assets. Applying big
data analytics design methods can provide potential values
in the transportation field in terms of cost, risk reduction,
delay reduction and optimization. A well-developed
evidence-based database is very helpful to transportation
executives and researchers as it is more strategic by taking
advantage of proven and useful information. Utilization of
evidence-based approaches with risk management requires
a great use of condition assessment data. This can help in
the identification and measurement of benefits achieved
with the optimization of the asset under management. One
of the key tasks confronting the asset management
custodian is the selection of the approach for big data
analytics techniques for classification of condition data.
Approaches utilizing big data analytics with design data
combined with organic data from measurements during the
asset lifecycle may be required including a selection of big
data technologies discussed earlier. Asset integrity reports
utilizing asset condition data are required for compliance

This work is licensed under a Creative Commons Attribution 4.0 License. For more information, see https://creativecommons.org/licenses/by/4.0/.
This article has been accepted for publication in a future issue of this journal, but has not been fully edited. Content may change prior to final publication. Citation information: DOI
10.1109/ACCESS.2020.2967436, IEEE Access
18
VOLUME 8, 2020
with the relevant standards or regulators’ requirements
within Australia and overseas. Current approaches do not
provide the required accuracy levels for classification of
condition monitoring data where there are several asset
condition classes. Hence further development of
autonomous approaches as outlined using artificial
intelligence may assist in meeting the required accuracy
levels required by organizations for safety-critical assets.
B. THE FUTURE OF BIG DATA IN INFRASTRUCTURE
ASSET MANAGEMENT
A key challenge facing railway asset owners and operators
is in being able to view the interdependencies of various
systems that are a part of the railway network and identify
root causes for failures and system behaviors before delays
occur [200]. This may also involve big data mining with
spatial and Spatio-temporal properties to provide
visualization of trends and relationships between assets.
The spatial dimension would involve the real-time location
of key assets and their respective systems and sub-systems
in each key asset. Technologies exist to provide the real-
time location information, but generally are stored
separately to measurement and log data. The temporal
dimension would involve the display of historical and
current asset condition measurements with performance
criteria as a sub-category. The temporal dimension would
require the integration of design and organic data to a
common asset data model with machine learning techniques
utilized to identify the relationships. While there would be
value in displaying asset performance in the two
dimensions, namely spatial and temporal, the further value
could be derived if a third dimension could be added to
show the interdependencies and hierarchy of various
systems that make up the network. The third dimension
would also involve artificial intelligence with expert
systems and machine learning utilized to identify the
interdependencies and hierarchies. While building
information management (BIM) systems exist for fixed
assets, a key challenge will be in integrating the fixed asset
information with mobile asset information using
heterogeneous data sources. The development of scenario-
based modeling with PHM would also become a necessary
adjunct to big data analytics in railway asset management.
C. RECOMMENDED RESEARCH DIRECTIONS IN USING
BIG DATA ANALYTICS
Research efforts are required to investigate and develop
commonly agreed processes and methods for application of
big data analytics as well as well accepted tools for
integration of asset data including condition monitoring
data in railway asset management systems.
Big data analytics technologies are still in their infancy
within the IT pillars of rail organizations. It is
recommended that the research efforts could be directed to
define potential big data analytics frameworks, and to
integrate different big data approaches for condition and
failure monitoring.
The introduction of big data analytics to the railway
sector will require further development of tools and models
for investigation of asset condition status such as using
multi-state degradation modeling and correlation analysis
of management activities in asset lifecycle.
Application of Blockchain technology in the railway
industry is at the start-up stage but the research direction is
promising. The future research efforts are directed at
developing new frameworks and standards for the industry
to use and easily implement their Blockchains. New
technologies will be explored to handle the issues related to
the required speed and scalability when developing
Blockchains for a complicated and busy network.
On the basis of the work presented in this paper, the
next step would be directed at providing a review of
application of big data analytics to PHM for railway
communication networks, infrastructure and transportation
systems.
REFERENCES
[1] M. Cox and D. Ellsworth, "Managing big data for scientific
visualization," ACM Siggraph, vol. 97, pp. 2138, August 1997.
[2] J. R. Mashey, "Big data and the next wave of infrastress," Computer
Science Division Seminar, University of California, Berkeley,
October 1997.
[3] S. Bryson, D. Kenwright, M. Cox, D. Ellsworth, and R. Haimes,
"Visually exploring gigabyte data sets in real-time,"
Communications of the ACM, vol. 42, no. 8, pp. 82-90, 1999, doi:
10.1145/310930.310977.
[4] T. J. Berners-Lee. "Information management: A proposal."
http://cds.cern.ch/record/369245/files/dd-89-001.pdf (accessed
April 2019).
[5] CERN Website. https://home.cern/science/computing/birth-
web/short-history-web (accessed April, 2019).
[6] F. Diebold, 2012. On the origins and development of Big Data: the
phenomenon, the term, and the discipline. línea. Disponible en<
https://economics. sas. upenn. edu/sites/economics. sas. upenn.
edu/files/12-037. pdf>[última consulta: 16 de marzo de 2016]. "On
the origins and development of Big Data: the phenomenon, the
term, and the discipline," 2019. [Online]. Available:
https://economics. sas. upenn. edu/sites/economics. sas. upenn.
edu/files/12-037. pdf>[última consulta: 16 de marzo de 2016].
[7] C. Chen and C. Zhang, "Data-intensive applications, challenges,
techniques and technologies: A survey on Big Data," Information
Sciences, vol. 275, pp. 314347, 2014, doi:
10.1016/j.ins.2014.01.015.
[8] I. A. T. Hashem, I. Yaqoob, N. B. Anuar, S. Mokhtar, A. Gani, and
S. Ullah Khan, "The rise of “big data” on cloud computing: Review
and open research issues," Information Systems, vol. 47, pp.
98115, 2015, doi: 10.1016/j.is.2014.07.006.
[9] A. Siddiqa, A. Karim, and A. Gani, "Big data storage technologies:
a survey," Frontiers of Information Technology & Electronic
Engineering, vol. 18, no. 8, pp. 10401070, 2017, doi:
10.1631/fitee.1500441.
[10] P. Grover and A. K. Kar, "Big Data Analytics: A Review on
Theoretical Contributions and Tools Used in Literature," Global
Journal of Flexible Systems Management, vol. 18, no. 3, pp.
203229, 2017, doi: 10.1007/s40171-017-0159-3.
[11] A. Oussous, F.-Z. Benjelloun, A. Ait Lahcen, and S. Belfkih, "Big
Data technologies: A survey," Journal of King Saud University -
Computer and Information Sciences, 2017, doi:
10.1016/j.jksuci.2017.06.001.
[12] A. Kamilaris, A. Kartakoullis, and F. X. Prenafeta-Boldú, "A review
on the practice of big data analysis in agriculture," Computers and
Electronics in Agriculture, vol. 143, pp. 2337, 2017, doi:
10.1016/j.compag.2017.09.037.

This work is licensed under a Creative Commons Attribution 4.0 License. For more information, see https://creativecommons.org/licenses/by/4.0/.
This article has been accepted for publication in a future issue of this journal, but has not been fully edited. Content may change prior to final publication. Citation information: DOI
10.1109/ACCESS.2020.2967436, IEEE Access
19
VOLUME 8, 2020
[13] J. Luo, M. Wu, D. Gopukumar, and Y. Zhao, "Big Data Application
in Biomedical Research and Health Care: A Literature Review,"
Biomed Inform Insights, vol. 8, pp. 110, 2016, doi:
10.4137/BII.S31559.
[14] S. Tiwari, H. M. Wee, and Y. Daryanto, "Big data analytics in
supply chain management between 2010 and 2016: Insights to
industries," Computers & Industrial Engineering, vol. 115, pp.
319330, 2018, doi: 10.1016/j.cie.2017.11.017.
[15] K. Nagorny, P. Lima-Monteiro, J. Barata, and A. Walter Colombo,
"Big data analysis in smart manufacturing: A review," International
Journal of Communications, Network and System Sciences, vol. 10,
pp. 3158, 2017, doi: 10.4236/ijcns.2017.103003.
[16] F. Ghofrani, Q. He, R. M. P. Goverde, and X. Liu, "Recent
applications of big data analytics in railway transportation systems:
A survey," Transportation Research Part C: Emerging
Technologies, vol. 90, pp. 226246, 2018, doi:
10.1016/j.trc.2018.03.010.
[17] H. J. Parkinson and G. Bamford, "The potential for using big data
analytics to predict safety risks by analyzing rail accidents,"
presented at the 3rd International Conference on Railway
Technology: Research, Development and Maintenance, Cagliari,
Sardinia, Italy, April, 2016.
[18] J. R. Kennedy and J. D. Parrish, "Total System Support Concept for
Railway Signalling Systems," presented at the Railway Engineering
Conference, Adelaide, SA, 2325 September 1991.
[19] ISO 55002 Asset management — Management systems —
Guidelines for the application of ISO 55001, International Standards
Organization, 2014.
[20] P. Scarf, R. Dwight, A. McCusker, and A. Chan, "Asset
replacement for an urban railway using a modified two-cycle
replacement model," Journal of the Operational Research Society,
vol. 58, no. 9, pp. 11231137, 2007, doi:
10.1057/palgrave.jors.2602288.
[21] R. Oweis, "Aligning Asset Management Policy, strategies and
proceses to deliver effective and sustainable results," presented at
the ICOMS Asset Management Conference, Gold Coast, QLD,
Australia 16-20 May, 2011.
[22] K. J. Epp and P. J. Mutton, "Wheel/Rail Interaction: Current'State of
the Art'in the Australasian Railway Industry.," presented at the
CORE 2002: Cost Efficient Railways through Engineering,
Wollongong, AU, 0-13 November, 2002.
[23] M. Palo, D. Galar, T. Nordmark, M. Asplund, and D. Larsson,
"Condition monitoring at the wheel/rail interface for decision-
making support," Proceedings of the Institution of Mechanical
Engineers, Part F: Journal of Rail and Rapid Transit, vol. 228, no. 6,
pp. 705-715, 2014, doi: 10.1177/0954409714526164.
[24] C. A. Lengnick-Hall and R. J. Griffith, "Evidence-based versus
tinkerable knowledge as strategic assets: A new perspective on the
interplay between innovation and application," Journal of
Engineering and Technology Management, vol. 28, no. 3, pp.
147167, 2011, doi: 10.1016/j.jengtecman.2011.03.003.
[25] D. S. Michele and L. Daniela, "Decision-support tools for municipal
infrastructure maintenance management," Procedia Computer
Science, vol. 3, pp. 36-41, 2011, doi: 10.1016/j.procs.2010.12.007.
[26] S. Jovanovic, "Modern railway infrastructure asset management,"
presented at the Proceedings of the 24th Southern African Transport
Conference (SATC 2005), Pretoria, South Africa, 11- 13 July, 2005.
[27] A. Thaduri, D. Galar, and U. Kumar, "Railway assets: A potential
domain for big data analytics," Procedia Computer Science, vol. 53,
pp. 457467, 2015, doi: 10.1016/j.procs.2015.07.323.
[28] C. D. Cottrill and S. Derrible, "Leveraging big data for the
development of transport sustainability indicators," Journal of
Urban Technology, vol. 22, no. 1, pp. 4564, 2015, doi:
10.1080/10630732.2014.942094.
[29] E. W. T. Ngai, Y. Hu, Y. H. Wong, Y. Chen, and X. Sun, "The
application of data mining techniques in financial fraud detection: A
classification framework and an academic review of literature,"
Decision Support Systems, vol. 50, no. 3, pp. 559569, 2011, doi:
10.1016/j.dss.2010.08.006.
[30] W. A. Günther, M. H. Rezazade Mehrizi, M. Huysman, and F.
Feldberg, "Debating big data: A literature review on realizing value
from big data," The Journal of Strategic Information Systems, vol.
26, no. 3, pp. 191209, 2017, doi: 10.1016/j.jsis.2017.07.003.
[31] IBM. "The four V’s of big data."
http://www.ibmbigdatahub.com/infographic/four-vs-big-data
(accessed June 2019).
[32] A. Gandomi and M. Haider, "Beyond the hype: Big data concepts,
methods, and analytics," International Journal of Information
Management, vol. 35, no. 2, pp. 137144, 2015, doi:
10.1016/j.ijinfomgt.2014.10.007.
[33] H. Wang, Z. Xu, H. Fujita, and S. Liu, "Towards felicitous decision
making: An overview on challenges and trends of Big Data,"
Information Sciences, vol. 367-368, pp. 747765, 2016, doi:
10.1016/j.ins.2016.07.007.
[34] L. Osuszek, S. Stanek, and Z. Twardowski, "Leverage big data
analytics for dynamic informed decisions with advanced case
management," Journal of Decision Systems, vol. 25, no. sup1, pp.
436449, 2016, doi: 10.1080/12460125.2016.1187401.
[35] A. L'Heureux, K. Grolinger, H. F. Elyamany, and M. A. M. Capretz,
"Machine Learning With Big Data: Challenges and Approaches,"
IEEE Access, vol. 5, pp. 77767797, 2017, doi:
10.1109/access.2017.2696365.
[36] Q. P. He and J. Wang, "Statistical process monitoring as a big data
analytics tool for smart manufacturing," Journal of Process Control,
vol. 67, pp. 3543, 2018, doi: 10.1016/j.jprocont.2017.06.012.
[37] E. P. Xing, Q. Ho, P. Xie, and D. Wei, "Strategies and Principles of
Distributed Machine Learning on Big Data," Engineering, vol. 2,
no. 2, pp. 179195, 2016, doi: 10.1016/j.Eng.2016.02.008.
[38] F. Portela, L. Lima, and M. F. Santos, "Why Big Data? Towards a
Project Assessment Framework," Procedia Computer Science, vol.
98, pp. 604609, 2016, doi: 10.1016/j.procs.2016.09.094.
[39] B. Franke et al., "Statistical inference, learning and models in big
data," International Statistical Review, vol. 84, no. 3, pp. 371389,
2016.
[40] M. Kans and D. Galar, "The Impact of Maintenance 4.0 and Big
Data Analytics within Strategic Asset Management," presented at
the 6th International Conference on Maintenance Performance
Measurement and Management, Lule, Sweden, 28 November,
2016.
[41] I. Lee, "Big data: Dimensions, evolution, impacts, and challenges,"
Business Horizons, vol. 60, no. 3, pp. 293303, 2017, doi:
10.1016/j.bushor.2017.01.004.
[42] S. Landset, T. M. Khoshgoftaar, A. N. Richter, and T. Hasanin, "A
survey of open source tools for machine learning with big data in
the Hadoop ecosystem," Journal of Big Data, vol. 2, no. 1, 2015,
doi: 10.1186/s40537-015-0032-1.
[43] A. Bechini, F. Marcelloni, and A. Segatori, "A MapReduce solution
for associative classification of big data," Information Sciences, vol.
332, pp. 3355, 2016, doi: 10.1016/j.ins.2015.10.041.
[44] J. Fan, F. Han, and H. Liu, "Challenges of Big Data Analysis," Natl
Sci Rev, vol. 1, no. 2, pp. 293314, Jun 2014, doi:
10.1093/nsr/nwt032.
[45] V. López, S. del Río, J. M. Benítez, and F. Herrera, "Cost-sensitive
linguistic fuzzy rule based classification systems under the
MapReduce framework for imbalanced big data," Fuzzy Sets and
Systems, vol. 258, pp. 538, 2015, doi: 10.1016/j.fss.2014.01.015.
[46] N. Khan, M. Alsaqer, H. Shah, G. Badsha, A. A. Abbasi, and S.
Salehian, "The 10 Vs, Issues and Challenges of Big Data,"
presented at the Proceedings of the 2018 International Conference
on Big Data and Education - ICBDE '18, Honolulu, HI, USA, 9-11
March, 2018.
[47] U. Sivarajah, M. M. Kamal, Z. Irani, and V. Weerakkody, "Critical
analysis of Big Data challenges and analytical methods," Journal of
Business Research, vol. 70, pp. 263286, 2017, doi:
10.1016/j.jbusres.2016.08.001.
[48] H. Watson, "Tutorial: Big Data Analytics: Concepts, Technologies,
and Applications," Communications of the Association for
Information Systems, vol. 34, no. 6, 2014.
[49] P. Raj, A. Raman, D. Nagaraj, and S. Duggirala, "In-Database
Processing and In-Memory Analytics," in High-Performance Big-
Data Analytics, (Computer Communications and Networks, 2015,
ch. Chapter 8, pp. 207231.

This work is licensed under a Creative Commons Attribution 4.0 License. For more information, see https://creativecommons.org/licenses/by/4.0/.
This article has been accepted for publication in a future issue of this journal, but has not been fully edited. Content may change prior to final publication. Citation information: DOI
10.1109/ACCESS.2020.2967436, IEEE Access
20
VOLUME 8, 2020
[50] H. Shu, "Big data analytics: six techniques," Geo-spatial
Information Science, vol. 19, no. 2, pp. 119128, 2016, doi:
10.1080/10095020.2016.1182307.
[51] D. den Hertog and K. Postek, "Bridging the gap between predictive
and prescriptive analytics-new optimization methodology needed,"
Tilburg University, Netherlands, 2016. [Online]. Available:
http://www. optimization-online. org/DB_HTML/2016/12/5779.
html.
[52] R. M. Groves, "Three Eras of Survey Research," Public Opinion
Quarterly, vol. 75, no. 5, pp. 861871, 2011, doi:
10.1093/poq/nfr057.
[53] P. Pace, l. Kontokostas , l. Wessels , and G. Sockalingam,
"Condition-based maintenance using an integrated track
maintenance management system," presented at the Conference On
Railway Engineering, Perth, WA, Australia, 7-10 September, 2008.
[54] D. Jagan, C. Southern, A. Kane, and K. Bladon, "Managing the
Australian defined interstate rail network wayside condition
monitoring data," presented at the CORE 2014: Rail Transport For
A Vital Economy., Adelaide, SA, 2014.
[55] K. Bladon, D. Rennison, R. Tracy, and T. Bladon, "Predictive
condition monitoring of railway rolling stock," presented at the
Conference On Railway Engineering, Darwin, NT, Australia, 20-23
June, 2004.
[56] P. R. Cohen, N. M. Adams, and M. Berthold, "Advances in
intelligent data analysis IX," in Proceedings of the 9th international
symposium, IDA 2010, Tucson, AZ, USA, May 19-21 2010.
[57] G. Hardie, G. Tew, G. Crew, and M. Courtney, "Track condition
assessment and monitoring in heavy haul railroads," presented at the
The 2011 International Heavy Haul Association Conference,
Calgary, Alberta, Canada, 19-22 June 2011.
[58] S. Doyle, C. Bastucescu, and T. Vale, "Pantograph condition
monitoring system for automated maintenance inspections and
prevention of overhead wiring tear downs," presented at the CORE
2016: Maintaining the Momentum, Melbourne, Australia, 16-18
May, 2016.
[59] K. Laakso, T. Rosqvist, and J. L. Paulsen, "The use of condition
monitoring information for maintenance planning and decision-
making," Nordisk Kernesikkerhedsforskning, 2002.
[60] M. Bengtsson, "Condition Based Maintenance on Rail Vehicles-
Possibilities for a more effective maintenance strategy," 2003.
[61] K. Lee, J. Lee, and I. Kim, "A study on strategy of condition based
maintenance for Korea metro rolling stocks," presented at the 7th
IET Conference on Railway Condition Monitoring 2016 (RCM
2016), Birmingham, UK, 27 - 28 September, 2016.
[62] S. Kalay and M. Witte, "Recent advances in automated wagon
health monitoring and its effects on safety in North America,"
presented at the IHHA 2015, Perth, WA, Australia, 21-24 June,
2015.
[63] K. Yano, T. Suzuki, K. Okada, W. Wang, and T. Takayanagi, "Data
modeling technology in railway operation and maintenance,"
Hitachi Review, 2018.
[64] D. Wobschall, "IEEE 1451—a universal transducer protocol
standard," presented at the 2007 IEEE Autotestcon, Baltimore, MD,
USA., 17-20 Sept, 2007.
[65] R. Nappi, "Integrated Maintenance: analysis and perspective of
innovation in railway sector " arXiv preprint arXiv:1404.7560.,
2014.
[66] M. Bengtsson, "Standardization issues in condition based
maintenance," presented at the Condition Monitoring and
Diagnostic Engineering Management - Proceedings of the 16th
International Congress, Vxj University, Sweden, 27-29 August
2003.
[67] S. Burnett and P. J. Vlok, "A simplified numerical decision-making
methodology for physical asset management decisions," South
African Journal of Industrial Engineering, vol. 25, no. 1, pp.
162175, May 2014.
[68] A. D. Mathew, S. Zhang, L. Ma, and D. Hargreaves, "A water
utility industry conceptual asset management data warehouse
model," presented at the The 36th International Conference on
Computers and Industrial Engineering, Howard Hotel, Taipei,
Taiwan., 2006.
[69] J. Tutcher, J. Easton, C. Roberts, R. Myall, M. Hargreaves, and C.
Tiller, "Ontology-based data management for the GB rail industry
Feasibility study," Rail Safety and Standards Board, London, UK,
2013.
[70] S. Liu, A. H. B. Duffy, R. I. Whitfield, and I. M. Boyle, "Integration
of decision support systems to improve decision support
performance," Knowledge and Information Systems, vol. 22, no. 3,
pp. 261286, 2009, doi: 10.1007/s10115-009-0192-4.
[71] W. W. Eckerson, "Deploying Dashboards and scorecards," TDWI
Best Practices Report, July 2006.
[72] A. Makris, K. Tserpes, V. Andronikou, and D. Anagnostopoulos,
"A Classification of NoSQL Data Stores Based on Key Design
Characteristics," Procedia Computer Science, vol. 97, pp. 94103,
2016, doi: 10.1016/j.procs.2016.08.284.
[73] B. E. James and P. O. Asagba, "Hybrid database system for Big
Data storage and management," International Journal of Computer
Science, Engineering and Applications, vol. 7, no. 3/4, pp. 1527,
2017, doi: 10.5121/ijcsea.2017.7402.
[74] P. Kecman and R. M. P. Goverde, "Process mining of train
describer event data and automatic conflict identification,"
presented at the Computers in Railways XIII, Southampton, UK,
2012.
[75] PAS 1192:2013 Part 2: Specification for information management
for the capital/delivery phase of construction projects using
building information modeling, BSI, 2013.
[76] P. Fraga-Lamas, T. M. Fernandez-Carames, and L. Castedo,
"Towards the internet of smart trains: A review on industrial IoT-
connected railways," Sensors, vol. 17, no. 6, pp. 14571501, Jun 21
2017, doi: 10.3390/s17061457.
[77] A. Borrmann, M. Hochmuth, M. König, T. Liebich, and D. Singer,
"Germany’s governmental BIM initiative–Assessing the
performance of the BIM pilot projects," presented at the 16th
International Conference on Computing in Civil and Building
Engineering, Osaka, Japan, 6-8 July, 2016.
[78] G. A. Boyes, C. Ellul, and D. Irwin, "Exploring bim for operational
integrated asset management - A preliminary study utilising real-
world infrastructure data," ISPRS Annals of Photogrammetry,
Remote Sensing and Spatial Information Sciences, vol. IV-4/W5,
pp. 4956, 2017, doi: 10.5194/isprs-annals-IV-4-W5-49-2017.
[79] A. Bradley, H. Li, R. Lark, and S. Dunn, "BIM for infrastructure:
An overall review and constructor perspective," Automation in
Construction, vol. 71, pp. 139152, 2016, doi:
10.1016/j.autcon.2016.08.019.
[80] A. Sanchez, J. A. Kraatz, K. D. Hampson, and S. Loganathan, "BIM
for sustainable whole-of-life transport infrastructure asset
management," presented at the Sustainability in Public Works
Conference, Tweed Heads, Australia, 7-29 July, 2014.
[81] G. Hao, S. Ma, J. Lv, and Y. Sui, "A service-oriented data
integration architecture and the integrating tree," presented at the
Proceedings of the Fifth International Conference on Grid and
Cooperative Computing (GCC'06), Hunan, China, 21 Oct - 23 Oct,
2006.
[82] J. Gao, A. P. Koronios, and C. S. Lin, "A data quality model for
asset management in engineering organizations," presented at the
Proceedings of the 2005 International Conference on Information
Quality, ICIQ 2005, MIT, Cambridge, Massachusetts, USA, 4 Nov -
6 Nov, 2005.
[83] A. Zoeteman, "Asset maintenance management: state of the art in
the European railways," International Journal of Critical
Infrastructures, vol. 2, no. 2-3, pp. 171186, 2006.
[84] V. Saquicela, V.-B. L.M., and A. Tello, "Challenges and Trends
about Smart Big Geospatial Data: A Position Paper," presented at
the IEEE International Conference on Big Data (BIGDATA),
Boston, MA, USA, 11-14 December, 2017.
[85] J. Whyte, A. Stasis, and C. Lindkvist, "Managing change in the
delivery of complex projects: Configuration management, asset
information and ‘big data’," International Journal of Project
Management, vol. 34, no. 2, pp. 339351, February 29 2016, doi:
10.1016/j.ijproman.2015.02.006.
[86] M. Takikawa, "Innovation in Railway Maintenance utilizing
Information and Communication Technology (Smart Maintenance

This work is licensed under a Creative Commons Attribution 4.0 License. For more information, see https://creativecommons.org/licenses/by/4.0/.
This article has been accepted for publication in a future issue of this journal, but has not been fully edited. Content may change prior to final publication. Citation information: DOI
10.1109/ACCESS.2020.2967436, IEEE Access
21
VOLUME 8, 2020
Initiative)," Communication Technology (Smart Maintenance
Initiative), pp. 2235, 2016.
[87] A. Alharthi, V. Krotov, and M. Bowman, "Addressing barriers to
big data," Business Horizons, vol. 60, no. 3, pp. 285292, 2017, doi:
10.1016/j.bushor.2017.01.002.
[88] P. J. H. Mooren and L. A. van Dongen, "Application of Remote
Condition Monitoring in Different Rolling Stock Life Cycle
Phases," Procedia CIRP, vol. 11, pp. 135138, 2013, doi:
10.1016/j.procir.2013.07.050.
[89] V. Rajaraman, "Big Data Analytics," Resonance, vol. 21, pp.
695716, 2016, doi: https://doi.org/10.1007/s12045-016-0376-7.
[90] Q. Chen, M. Hsu, and H. Zeller, "Experience in Continuous
analytics as a Service (CaaaS)," presented at the 14th International
Conference on Extending Database Technology, Uppsala, Sweden,
21-24 March 2011.
[91] C. Evangelinos, P. F. J. Lermusiaux, X. Jinshan, P. J. Haley, and C.
N. Hill, "Many Task Computing for Real-Time Uncertainty
Prediction and Data Assimilation in the Ocean," IEEE Transactions
on Parallel and Distributed Systems, vol. 22, no. 6, pp. 10121024,
2011, doi: 10.1109/tpds.2011.64.
[92] G. Manco et al., "Fault detection and explanation through big data
analysis on sensor streams," Expert Systems with Applications, vol.
87, pp. 141156, 2017.
[93] D. E. Boyle, D. C. Yates, and E. M. Yeatman, "Urban sensor data
streams," EEE Internet Computing, vol. 17, no. 6, pp. 1220, 2013.
[94] R. Taverner and V. Lammerse, "Implementing Maximo Asset
Management System for Adelaide's public transport rail assets,"
presented at the CORE 2014: Rail Transport For A Vital Economy,
Adelaide, SA, Australia, 5-7 May, 2014.
[95] S. Shekhar, P. Zhang, and Y. Huang, "Spatial Data Mining," in
Spatial data mining. Data Mining and Knowledge Discovery
Handbook, O. Z. Mainmon and L. Rokach Eds. New York:
Springer, USA, 2005, pp. 833851.
[96] H. Chen, R. H. Chiang, and V. C. Storey, "Business intelligence and
analytics: From big data to big impact," MIS quarterly, vol. 36, no.
4, 2012.
[97] E. E. Mangina, S. D. J. McArthur, and J. R. McDonald,
"Autonomous Agents for Distributed Problem Solving in Condition
Monitoring," in Intelligent Problem Solving. Methodologies and
Approaches, vol. 1821, (Lecture Notes in Computer Science. Berlin
Heidelberg: Springer, 2000, pp. 683692.
[98] E. Fumeo, L. Oneto, and D. Anguita, "Condition based maintenance
in railway transportation systems based on big data streaming
analysis," Procedia Computer Science, vol. 53, pp. 437446, 2015,
doi: 10.1016/j.procs.2015.07.321.
[99] M. L. Brown and J. F. Kros, "Data mining and the impact of
missing data," Industrial Management & Data Systems, vol. 103,
no. 8, pp. 611621, 2003.
[100] P. Leonard, "Data and the internet of things," presented at the AFR
Innovation Summit, Sydney, Australia, 16 - 18 August, 2016.
[101] P. E. D. Love, J. Zhou, J. Matthews, M. Lavender, and T. Morse,
"Managing rail infrastructure for a digital future: Future-proofing of
asset information," Transportation Research Part A: Policy and
Practice, vol. 110, pp. 161176, 2018, doi:
10.1016/j.tra.2018.02.014.
[102] A. D’agostino, "Big data in railways Common Occurrence
Reporting Programme," European Union Agency for Railways,
2016.
[103] D. Karamshuk, F. Shaw, J. Brownlie, and N. Sastry, "Bridging big
data and qualitative methods in the social sciences: A case study of
Twitter responses to high profile deaths by suicide," Online Social
Networks and Media, vol. 1, pp. 3343, 2017, doi:
10.1016/j.osnem.2017.01.002.
[104] R. M. Kaplan, D. A. Chambers, and R. E. Glasgow, "Big data and
large sample size: a cautionary note on the potential for bias," Clin
Transl Sci, vol. 7, no. 4, pp. 342-6, Aug 2014, doi:
10.1111/cts.12178.
[105] D. B. Dunson, "Statistics in the big data era: Failures of the
machine," Statistics & Probability Letters, vol. 136, pp. 49, 2018,
doi: 10.1016/j.spl.2018.02.028.
[106] J. Thomas and L. Sael, "Overview of integrative analysis methods
for heterogeneous data," presented at the International Conference
on Big Data and Smart Computing (BigComp), Jeju Island, Korea,
9-12 February, 2015.
[107] V. Subramaniyaswamy, V. Vijayakumar, R. Logesh, and V.
Indragandhi, "Unstructured data analysis on big data Using Map
Reduce," Procedia Computer Science, vol. 50, pp. 456465, 2015,
doi: 10.1016/j.procs.2015.04.015.
[108] S. Kaisler, F. Armour, J. A. Espinosa, and W. Money, "Big Data:
Issues and Challenges Moving Forward," presented at the 46th
Hawaii International Conference on System Sciences, Wailea,
Maui, HI USA, 7-10 January, 2013.
[109] National Research Council, Reliability Growth: Enhancing Defense
System Reliability. Washington, DC, USA: National Academies
Press, 2015.
[110] S. D. Griffith, D. J. Quest, T. S. Brettin, and R. W. Cottingham,
"Scenario driven data modeling: a method for integrating diverse
sources of data and data streams," BMC Bioinformatics, vol. 12
Suppl 10, p. S17, Oct 18 2011, doi: 10.1186/1471-2105-12-S10-
S17.
[111] F. Gao, S. Bhiri, and Z. Zhou, "User-centric modeling and
processing for ubiquitous events using semantic capability models,"
Communications in Mobile Computing, vol. 1, no. 1, pp. 17, 2012.
[112] B. Eine, M. Jurisch, and W. Quint, "Ontology-Based Big Data
Management," Systems, vol. 5, no. 3, 2017, doi:
10.3390/systems5030045.
[113] O. Brazhnik and J. F. Jones, "Anatomy of data integration," Journal
of Biomedical Informatics, vol. 40, no. 3, pp. 252–269, 2007, doi:
http://doi.org/10.1016/j.jbi.2006.09.001. (accessed May, 2019).
[114] H. Shang and C. Berenguer, "A Colored Petri Net model for railway
track maintenance with two-level inspection," presented at the
Proceedings of European Safety and Reliability Conference
(ESREL 2014), Wroclaw, Poland, 14-18 September, 2014.
[115] Q. Li, Z. Zhong, Z. Liang, and L. Y., "Rail inspection meets big
data: Methods and trends," presented at the 2015 18th International
Conference on Network-Based Information Systems (NBiS), Taipei,
Taiwan, 2-4 September, 2015.
[116] A. A. da Silva, G. L. Bellotti, and H. H. Chaya, "Utilizing asset data
for predictive asset management," INTECH,
NOVEMBER/DECEMBER 2011.
[117] S. G. Alonso, I. de la Torre Díez, J. J. P. C. Rodrigues, S. Hamrioui,
and M. López-Coronado, "A Systematic Review of Techniques and
Sources of Big Data in the Healthcare Sector," Journal of Medical
Systems, vol. 41, no. 11, 2017, doi: 10.1007/s10916-017-0832-2.
[118] C. K. Emani, N. Cullot, and C. Nicolle, "Understandable Big Data:
A survey," Computer Science Review, vol. 17, pp. 7081, 2015, doi:
10.1016/j.cosrev.2015.05.002.
[119] H. Shuijing, "Big Data Analytics: Key Technologies and
Challenges," presented at the 2016 International Conference on
Robots & Intelligent System (ICRIS), ZhangJiaJie, China, 27-28
August, 2016.
[120] X. Zhou et al., "Big Data Storage and Parallel Analysis of Grid
Equipment Monitoring System," International Journal of
Performability Engineering, vol. 14, no. 2, pp. 202209, 2018, doi:
10.23940/ijpe.18.02.p2.202209.
[121] C. Lakshmi and V. V. Nagendra Kumar, "Survey Paper on Big
Data," International Journal of Advanced Research in Computer
Science and Software Engineering, vol. 6, no. 8, pp. 368381, 2016.
[122] N. Honest, "A Survey of Big Data Analytics," International Journal
of Information Sciences and Techniques, vol. 6, no. 1/2, pp. 3543,
2016, doi: 10.5121/ijist.2016.6204.
[123] M. Praveena and M. Kameswara Rao, "Survey on Big data analytics
in Healthcare Domain," International Journal of Engineering &
Technology, vol. 7, no. 2.7, pp. 919925, 2018.
[124] C.-W. Tsai, C.-F. Lai, H.-C. Chao, and A. V. Vasilakos, "Big data
analytics: a survey," Journal of Big Data, vol. 2, no. 1, 2015, doi:
10.1186/s40537-015-0030-3.
[125] H. Han, W. Yonggang, C. Tat-Seng, and L. Xuelong, "Toward
Scalable Systems for Big Data Analytics: A Technology Tutorial,"
IEEE Access, vol. 2, pp. 652687, 2014, doi:
10.1109/access.2014.2332453.

This work is licensed under a Creative Commons Attribution 4.0 License. For more information, see https://creativecommons.org/licenses/by/4.0/.
This article has been accepted for publication in a future issue of this journal, but has not been fully edited. Content may change prior to final publication. Citation information: DOI
10.1109/ACCESS.2020.2967436, IEEE Access
22
VOLUME 8, 2020
[126] A. Bahga and V. K. Madisetti, "Blockchain Platform for Industrial
Internet of Things," Journal of Software Engineering and
Applications, vol. 09, no. 10, pp. 533546, 2016, doi:
10.4236/jsea.2016.910036.
[127] Blockchain Hub, "Blockchains and Distributed Ledger
Technologies." [Online]. Available: https://blockchainhub.
net/blockchains-and-distributed-ledger-technologies-in-general/
[128] A. Gausdal, K. Czachorowski, and M. Solesvik, "Applying
Blockchain Technology: Evidence from Norwegian Companies,"
Sustainability, vol. 10, no. 6, 2018, doi: 10.3390/su10061985.
[129] N. Teslya and I. Ryabchikov, "Blockchain-based platforrm
architecture for industrial IoT," presented at the 21st Conference of
Open Innovations Association (FRUCT) Helsinki, Finland, 6-10
November, 2017.
[130] P. Ghuli, U. Priyam Kumar, and R. Shettar, "A review on
blockchain application for decentralized decision of ownership of
IoT devices," Advances in Computational Sciences and Technology,
vol. 10, no. 8, pp. 24492456, 2017.
[131] A. Dorri, S. S. Kanhere, and R. Jurdak, "Blockchain in internet of
things: challenges and solutions," arXiv preprint arXiv:1608.05187,
2016.
[132] A. Ouaddah, A. A. El Kalam, and A. A. Ouahman, "Harnessing the
power of blockchain technology to solve IoT security & privacy
issues," presented at the ICC 2017, Cambridge city, United
Kingdom, 22–23 March, 2017.
[133] M. Kuperberg, D. Kindler, and S. Jeschke, "Are Smart Contracts
and Blockchains Suitable for Decentralized Railway Control?,"
arXiv preprint arXiv:1901.06236, 2019.
[134] L. X. Downey, F. Bauchot, and J. Röling, "Blockchain for Business
Value: A Contract and Work Flow Management to Reduce Disputes
Pilot Project," IEEE Engineering Management Review, vol. 46, no.
4, pp. 8693, 2018.
[135] O. Probert. (2018, July 31st) Pacific National takes part in
experimental blockchain delivery. Rail Express.
[136] M. Andoni et al., "Blockchain technology in the energy sector: A
systematic review of challenges and opportunities," Renewable and
Sustainable Energy Reviews, vol. 100, pp. 143174, 2019, doi:
10.1016/j.rser.2018.10.014.
[137] A. Alexandre. "National Swiss Railway operator completes pilot of
Blockchain ID management system."
https://dollardestruction.com/18847/. (accessed May 2019).
[138] M. Rodríguez. "Swiss Federal Railways develop a digital identity
pilot on Ethereum." https://www.cryptoworldjournal.com/swiss-
federal-railways-develop-a-digital-identity-pilot-on-ethereum/.
(accessed May 2019).
[139] N. Newman. "How can Blockchain open up new opportunities for
rail freight." https://www.nicnewmanoxford.com/how-can-
blockchain-open-up-new-opportunities-for-rail-freight/. (accessed
May 2019).
[140] BiTA. "BiTA appoints IEEE-ISTO for standards."
https://www.freightwaves.com/news/blockchain/bita-to-use-ieee-
standards (accessed May 2019).
[141] F. Memoria. "Leading UK public transport provider to launch
Blockchain-based loyalty program." (accessed May 2019).
[142] S. Morant, "How can Blockchain open up new opportunities for rail
freight?," nternational Railway Journal, June 11 2018. [Online].
Available: https://www.railjournal.com/in_depth/how-can-
blockchain-open-up-new-opportunities-for-rail-freight.
[143] S. Bhatanagar. "Russian Railways are eyeing crypto for tickets,
Blockchain for cargo." http://blockchainmagnets.com/russian-
railways-are-eyeing-crypto-for-tickets/. (accessed May 2019).
[144] M. Popke, "Technology disruptor: Blockchain is a potential game-
changer in the rail realm," Rail News: Internet-Digital, April 2018.
[Online]. Available:
https://www.progressiverailroading.com/internet-
digital/article/Technology-disruptor-Blockchain-is-a-potential-
game-changer-in-the-rail-realm-industry-adopters-say--54395.
[145] O. Emmanuel, "Shenzhen adopts Blockchain technology for
transportation," Bitcoin, Blockchain & Cryptocurrency News, 20
March 2019. [Online]. Available: https://btcmanager.com/shenzhen-
adopts-blockchain-technology-transportation/?q=/shenzhen-adopts-
blockchain-technology-transportation/.
[146] M. H. Eiza, M. Randles, P. Johnson, N. Shone, J. Pang, and A.
Bhih, "Rail internet of things: An architectural platform and assured
requirements model," presented at the 2015 IEEE International
Conference on Computer and Information Technology; Ubiquitous
Computing and Communications; Dependable, Autonomic and
Secure Computing; Pervasive Intelligence and Computing,
Liverpool, UK,, 26-28 October, 2015.
[147] J. M. Easton, "Blockchains: A Distributed Data Ledger for the Rail
Industry," Innovative Applications of Big Data in the Railway
Industry, S. Kohli, A. Kumar, J. M. Easton, and C. Roberts, Eds.,
Hershey PA, USA GI Global, 2018, pp. 2739.
[148] A. Labrinidis and H. V. Jagadish, "Challenges and opportunities
with big data," in Proceedings of the VLDB Endowment, 2012, vol.
5, pp. 20322033.
[149] J. Dean and S. Ghemawat, "MapReduce: simplified data processing
on large clusters," Communications of the ACM, vol. 51, no. 1, pp.
107113, 2008.
[150] P. Louridas and C. Ebert, "Embedded analytics and statistics for big
data," IEEE software, vol. 30, no. 6, pp. 3339, November 2013.
[151] M. Mohammadi, A. Al-Fuqaha, S. Sorour, and M. Guizani, "Deep
Learning for IoT Big Data and Streaming Analytics: A Survey,"
arXiv preprint arXiv:1712.04301, 2017.
[152] Z. Li, H. Wang, W. Shao, J. Li, and H. Gao, "Repairing data
through regular expressions," presented at the Proceedings of the
VLDB Endowment, 2016.
[153] I. H. Witten, E. Frank, M. A. Hall, and C. J. Pal, Data mining:
practical machine learning tools and techniques, Sydney: Morgan
Kaufmann, 2016.
[154] M. U. Ali, S. Ahmad, and J. Ferzund, "Harnessing the potential of
machine learning for bioinformatics using big data tools,"
International Journal of Computer Science and Information
Security, vol. 14, no. 10, pp. 668675, October 2016.
[155] Y. Liu, L. Xu, and M. Li, "The Parallelization of Back Propagation
Neural Network in MapReduce and Spark," International Journal of
Parallel Programming, vol. 45, no. 4, pp. 760779, 2016, doi:
10.1007/s10766-016-0401-1.
[156] B. Liu, E. Blasch, Y. L. Chen, D. Shen, and G. Chen, "Scalable
sentiment classification for big data analysis using naive bayes
classifier," presented at the IEEE International Conference on Big
Data, Silicon Valley, CA, USA, 6-9 Oct. 2013.
[157] J. Khairnar and M. Kinikar, "Sentiment analysis based mining and
summarizing using SVM-MapReduce," International Journal of
Computer Science and Information Technologies, vol. 5, no. 3, pp.
40814085, 2014.
[158] N. Attoh-Okine, "Big data challenges in railway engineering,"
presented at the IEEE International Conference on Big Data (Big
Data), Hyatt Regency, Washington, DC, USA, 27 Oct-30 Oct, 2014.
[159] A. M. Zarembski, "Some examples of big data in railroad
engineering," presented at the IEEE International Conference on
Big Data, Washington, DC, USA, 27-30 Oct, 2014.
[160] S. F. Wamba, A. Gunasekaran, S. Akter, S. J.-f. Ren, R. Dubey, and
S. J. Childe, "Big data analytics and firm performance: Effects of
dynamic capabilities," Journal of Business Research, vol. 70, pp.
356365, 2017, doi: 10.1016/j.jbusres.2016.08.009.
[161] E. Raguseo, "Big data technologies: An empirical investigation on
their adoption, benefits and risks for companies," International
Journal of Information Management, vol. 38, no. 1, pp. 187195,
2018, doi: 10.1016/j.ijinfomgt.2017.07.008.
[162] S. Venkatraman, K. F. S. Kaspi, and R. Venkatraman, "SQL versus
NoSQL movement with Big Data Analytics," International Journal
of Information Technology and Computer Science, vol. 8, no. 12,
pp. 5966, 2016, doi: 10.5815/ijitcs.2016.12.07.
[163] A. Malviya, A. Udhani, and S. Soni, "R-Tool: Data analytic
framework for big data," presented at the Symposium on Colossal
Data Analysis and Networking (CDAN), Indore, Madhya Pradesh ,
India, 8-19 March, 2016.
[164] I. Durazo-Cardenas et al., "An autonomous system for maintenance
scheduling data-rich complex infrastructure: Fusing the railways’
condition, planning and cost," Transportation Research Part C:
Emerging Technologies, vol. 89, pp. 234253, 2018, doi:
10.1016/j.trc.2018.02.010.

This work is licensed under a Creative Commons Attribution 4.0 License. For more information, see https://creativecommons.org/licenses/by/4.0/.
This article has been accepted for publication in a future issue of this journal, but has not been fully edited. Content may change prior to final publication. Citation information: DOI
10.1109/ACCESS.2020.2967436, IEEE Access
23
VOLUME 8, 2020
[165] Deloitte. "Cloud-based data analytics enables Network Rail to
improve the performance of its network." Deloitte.
https://www2.deloitte.com/content/dam/Deloitte/uk/Documents/con
sultancy/deloitte-uk-network-rail.pdf (accessed 4th April 2019.
[166] L. Oneto et al., "Advanced Analytics for Train Delay Prediction
Systems by Including Exogenous Weather Data," presented at the
IEEE International Conference on Data Science and Advanced
Analytics (DSAA), Montreal Quebec, CA, 17-19 October, 2016.
[167] R. Furutani, F. Kudo, and N. Moriwaki, "Utilization of AI in the
Railway Sector: Case Study of Energy Efficiency in Railway
Operations," Hitachi Review, vol. 65, no. 6, pp. 128133, 2016.
[168] C. Chen, J. Ma, Y. Susilo, Y. Liu, and M. Wang, "The promises of
big data and small data for travel behavior (aka human mobility)
analysis," Transportation research part C: emerging technologies,
vol. 68, pp. 285299, Jul 2016, doi: 10.1016/j.trc.2016.04.005.
[169] A. Ø. Sørensen, N. O. E. Olsson, M. M. Akhtar, and H. Bull-Berg,
"Approaches, technologies and importance of analysis of the
number of train travellers," Urban, Planning and Transport
Research, vol. 7, no. 1, pp. 118, 2019, doi:
10.1080/21650020.2019.1566022.
[170] A. Tibaut, B. Kaučič, and D. Rebolj, "A standardised approac h for
sustainable interoperability between public transport passenger
information systems," Computers in Industry, vol. 63, no. 8, pp.
788798, 2012, doi: 10.1016/j.compind.2012.08.002.
[171] Manual of Standards and Recommended Practices Section A, Part
I, AAR, USA, 2019.
[172] G. Palem, "Designing condition-based maintenance management
systems for high-speed fleet," International Journal of Computer
Science and Business Informatics, vol. 17, no. 1, pp. 2840, 2017.
[173] I. Patel, A. Rau-Chaplin, and B. Varghese, "A Platform for Parallel
R-based Analytics on Cloud Infrastructure," presented at the 41st
International Conference on Parallel Processing Workshops,
Pittsburgh, PA, USA, 10-13 September, 2012.
[174] L. He, Q. Liang, and S. Fang, "Challenges and Innovative Solutions
in Urban Rail Transit Network Operations and Management:
China’s Guangzhou Metro Experience," Urban Rail Transit, vol. 2,
no. 1, pp. 3345, 2016, doi: 10.1007/s40864-016-0036-y.
[175] D. Bumblauskas, D. Gemmill, A. Igou, and J. Anzengruber, "Smart
Maintenance Decision Support Systems (SMDSS) based on
corporate big data analytics," Expert Systems with Applications, vol.
90, pp. 303317, 2017, doi: 10.1016/j.eswa.2017.08.025.
[176] S. K. Kinnunen, A. Ylä-Kujala, S. Marttonen-Arola, T. Kärri, and
D. Baglee, "Internet of things technologies to rationalize the data
acquisition in industrial asset management," presented at the 2016
World Congress in Computer Science, Computer Engineering, &
Applied Computing, Las Vegas, Nevada, USA, 25-28 July 2016.
[177] J. Lee, H. D. Ardakani, S. Yang, and B. Bagheri, "Industrial Big
Data Analytics and Cyber-physical Systems for Future Maintenance
& Service Innovation," Procedia CIRP, vol. 38, pp. 3-7, 2015, doi:
10.1016/j.procir.2015.08.026.
[178] Y. He, H. Tan, W. Luo, S. Feng, and J. Fan, "MR-DBSCAN: a
scalable MapReduce-based DBSCAN algorithm for heavily skewed
data," Frontiers of Computer Science, vol. 8, no. 1, pp. 8399,
2013, doi: 10.1007/s11704-013-3158-3.
[179] P. Pääkkönen and D. Pakkala, "Reference Architecture and
Classification of Technologies, Products and Services for Big Data
Systems," Big Data Research, vol. 2, no. 4, pp. 166186, 2015, doi:
10.1016/j.bdr.2015.01.001.
[180] X. Jin, B. W. Wah, X. Cheng, and Y. Wang, "Significance and
Challenges of Big Data Research," Big Data Research, vol. 2, no. 2,
pp. 5964, 2015, doi: 10.1016/j.bdr.2015.01.006.
[181] A. Núñez, J. Hendriks, Z. Li, B. De Schutter, and R. Dollevoet,
"Facilitating maintenance decisions on the Dutch railways using big
data: The ABA case study," presented at the IEEE International
Conference on Big Data, Washington, DC, USA, 27-30 October,
2014.
[182] V. Chandola, A. Banerjee, and V. Kumar, "Anomaly detection,"
ACM Computing Surveys, vol. 41, no. 3, pp. 158, 2009, doi:
10.1145/1541880.1541882.
[183] Y. Peng, M. Dong, and M. J. Zuo, "Current status of machine
prognostics in condition-based maintenance: a review," The
International Journal of Advanced Manufacturing Technology, vol.
50, no. 1-4, pp. 297313, 2010, doi: 10.1007/s00170-009-2482-0.
[184] D. Kwon, M. R. Hodkiewicz, J. Fan, T. Shibutani, and M. G. Pecht,
"IoT-Based Prognostics and Systems Health Management for
Industrial Applications," IEEE Access, vol. 4, pp. 3659-3670, 2016,
doi: 10.1109/access.2016.2587754.
[185] D. Zhang, C. Cadet, N. Yousfi-Steiner, F. Druart, and C. Bérenguer,
"PHM-oriented Degradation Indicators for Batteries and Fuel
Cells," Fuel Cells, vol. 17, no. 2, pp. 268276, 2017, doi:
10.1002/fuce.201600075.
[186] C. Kulkarni, J. R. Ceyla, G. Biswas, and K. Goebel, "Physics based
modeling and prognostics of electrolytic capacitors," American
Institute of Aeronautics and Astronautics, 2012.
[187] E. Zio, "Some Challenges and Opportunities in Reliability
Engineering," IEEE Transactions on Reliability, vol. 65, no. 4, pp.
17691782, 2016, doi: 10.1109/tr.2016.2591504.
[188] M. Schwabacher and K. Goebel, "A survey of artificial intelligence
for prognostics," presented at the Aaai fall symposium, 2007.
[189] G. W. Vogl, B. A. Weiss, and M. Helu, "A review of diagnostic and
prognostic capabilities and best practices for manufacturing,"
Journal of Intelligent Manufacturing, 2016, doi: 10.1007/s10845-
016-1228-8.
[190] D. An, J. H. Choi, and N. H. Kim, "Options for Prognostics
Methods: A review of data-driven and physics-based prognostics,"
presented at the 54th AIAA/ASME/ASCE/AHS/ASC Structures,
Structural Dynamics, and Materials Conference, Boston,
Massachusetts, USA., 8-11 April, 2013.
[191] P. Dersin, "Keynote speech: PHM in Railways: Big Data or Smart
Data?," presented at the Prognostics and System Health
Management Conference (PHM-Harbin), Harbin, China, 9-12 July,
2017.
[192] O. O. Aremu, A. Salvador Palau, D. Hyland-Wood, A. K. Parlikad,
and P. R. McAree, "Structuring Data for Intelligent Predictive
Maintenance in Asset Management," presented at the 16th IFAC
Symposium on Information Control Problems in Manufacturing.,
Bergamo, Italy, 11-13 June, 2018.
[193] A. N. Gorban and A. Y. Zinovyev, "Principal graphs and
manifolds," in Handbook of research on machine learning
applications and trends: algorithms, methods, and techniques IGI
Global, pp. 2859, 2010.
[194] M. Vignesh, "Predictive analytics in data mining with big data: A
literature survey," International Journal of Research in Engineering
and Applied Sciences (IJREAS), vol. 6, no. 11, pp. 1022, 2016.
[195] W. Sammouri, E. Come, L. Oukhellou, P. Aknin, and C. E.
Fonlladosa, "Floating train data systems for preventive
maintenance: A data mining approach," presented at the
International Conference on Industrial Engineering and Systems
Management (IESM), Rabat, Morocco, 28-30 October, 2013.
[196] B. Rapolu. "Bringing artificial intelligence to the rail industry "
Data Economy. http://www.techx365.com/document.asp?doc_id=
731414 (accessed January 2019).
[197] F. Cannarile, M. Compare, F. Di Maio, and E. Zio, "A clustering
approach for mining reliability big data for asset management,"
Proceedings of the Institution of Mechanical Engineers, Part O:
Journal of Risk and Reliability, vol. 232, no. 2, pp. 140150, 2018,
doi: 10.1177/1748006x17716344.
[198] A. Fahad et al., "A survey of clustering algorithms for big data:
Taxonomy and empirical analysis," IEEE transactions on emerging
topics in computing,, vol. 2, no. 3, pp. 267279, 2014.
[199] M. Menéndez, C. Martínez, G. Sanz, and J. M. Benitez,
"Development of a Smart Framework Based on Knowledge to
Support Infrastructure Maintenance Decisions in Railway
Corridors," Transportation Research Procedia, vol. 14, pp.
19871995, 2016, doi: 10.1016/j.trpro.2016.05.166.
[200] R. Veit-Egerer et al., "EINSTEIN Risk-Based Decision Model for
the Determination of Optimized Maintenance Intervention
Schedules for Infrastructure.," presented at the ABSE Conference
2018 – Engineering the Past, to Meet the Needs of the Future,
Copenhagen, Denmark, June 25-27, 2018.