Available online at www.sciencedirect.

com

ScienceDirect
Transportation Research Procedia 14 (2016) 1987 – 1995

6th Transport Research Arena April 18-21, 2016

Development of a smart framework based on knowledge to support

infrastructure maintenance decisions in railway corridors
Manuel Menéndez a,*, Carlos Martínez a, Gregorio Sanz b, Jose Manuel Benitez c
a
VIAS Y CONSTRUCCIONES S.A., Calle Orense nº11, Madrid 28020, Spain
b
FUNDACIÓN CARTIF, Parque Tecnológico Boecillo 205, Boecillo, 47151, Spain
c
UNIVERSIDAD DE GRANADA, E.T.S.I. Informática y Telecomunicación, Granada, 18071, Spain

Abstract

In a context of wide use of transport, it is necessary to increase the efficiency of different transport modes as well as their
interaction. To that effect, railway transport will play an important role in the future by increasing its capacity and ensuring
a sustainable, efficient and safe service. In that sense, it is essential to improve the interoperability and safety of national
networks in order to promote a single European Rail Market. On the other hand, the current railways systems need an efficient
maintenance in order to extend their lifetime.
Within this framework, the main objective of the OPTIRAIL project has been the development a set of new tools to support
smart maintenance decision making. These tools are based on Fuzzy and Computational Intelligence techniques. The tools are
integrated in a framework built out of railway infrastructure maintenance data and knowledge expertise across European railway
corridors. The framework has been evaluated through two case studies. The use of the system for the stakeholders will enable
better cross-border coordination for maintenance decision making. Furthermore, some non-participant railway administrators
have shown their interest and commitment to the project and its proposal.

©
© 2016 The Authors.
2016The Authors.Published
Publishedby by
Elsevier B.V.B.V..
Elsevier This is an open access article under the CC BY-NC-ND license
(http://creativecommons.org/licenses/by-nc-nd/4.0/).
Peer-review under responsibility of Road and Bridge Research Institute (IBDiM).
Peer-review under responsibility of Road and Bridge Research Institute (IBDiM)
Keywords: maintenance; railway; computational inteligence; fuzzy logic; corridors

* Corresponding author. Tel.: +34-914-179-800; fax: +34-914-179-833.

E-mail address: [email protected]

2352-1465 © 2016 The Authors. Published by Elsevier B.V. This is an open access article under the CC BY-NC-ND license
(http://creativecommons.org/licenses/by-nc-nd/4.0/).
Peer-review under responsibility of Road and Bridge Research Institute (IBDiM)
doi:10.1016/j.trpro.2016.05.166
1988 Manuel Menéndez et al. / Transportation Research Procedia 14 (2016) 1987 – 1995

1. Introduction

European transport networks have been growing over centuries as a consequence of the continuous increase of
volumes of passengers and goods. The existing networks have been straining beyond their capacity and, therefore,
there is a need to modernize the transport systems in order to allow networks to become the arteries and lifelines of
our society. Supporting the constantly increasing need for mobility, with no interruption of service, is the
fundamental prerequisite for Europe to ensure its future prosperity and territorial cohesion as well as to remain
a global level-playing actor under the pressure of the international market.
To that effect, the main changes in the European transport networks will focus on the available technology and
a fair use of transport modes. On the one hand, the level of maintenance operations has to be risen up and adapted
based on management tools and Information and Communication Technologies (ICT) systems. This will help to
reduce congestions, delays and accidents, which will lead to provide users and operators with predictable services.
On the other hand, there is a need to develop a European transport system capable of shifting the balance between
modes of transport, revitalising the railways, among others.
In the last 20 years, the European Commission has been very active in restructuring the European rail transport
market and strengthening the position of railways with respect to other transport modes. The Commission’s efforts
have concentrated on three major areas which are all crucial for developing a strong, transparent and competitive rail
transport industry: (i) opening of the rail transport to market competition, (ii) improving the interoperability and
safety of national networks and (iii) developing rail transport infrastructure.
Within this framework, maintenance of technical systems has become increasingly important in the rail sector.
Failures of systems may cause expensive production losses and can have negative effects on the people and
environment. Thus, maintenance is critical for ensuring safety, train punctuality, overall capacity utilization and
lower costs for modern railways.
Maintenance productivity is directly related to the available time-windows of train services and, therefore,
advanced management techniques and procedures get into action as a tool to combine the objective pursuit with both
needs: train service and maintenance. To that effect, different organisational entities must be set up for transport
operation and infrastructure management in order to co-operate and develop the infrastructure further.
Furthermore, improving the interoperability and safety of national networks is essential to promote a single
European Rail Market, but there are still some technical and non-technical barriers that need to be overcome for the
sake of interoperability.
Within the framework mentioned above the project “OPTIRAIL” is born.

Increase levels of reliability and safety

Railway services are characterized by a high degree of reliability and safety. The growing use of railway also
increases the need for maintenance, not only because of a higher degradation of the system, but also because the
availability of the track for maintenance decreases. A major constraint when organizing maintenance tasks is to
avoid disruption of service. Moreover, railway track maintenance is very little automated. This situation deserves
a close look in order to develop solutions that contribute to a more competitive and sustainable rail system.
To that effect, the OPTIRAIL project has developed a comprehensive framework, based on Fuzzy and
Computational Intelligent techniques, Klir, B. Yuan (1995) and Kruse and C. Borgelt (2013), to manage all the
elements that are relevant for track maintenance, predicting future conservations needs with optimal allocations of
resources.
This model has been developed based on the know-how of the organizations involved in the project. This know-
how has been extracted from expert knowledge and data on ICT systems, P. Pace and L. Kontokostas (2013).
All of this allows for a better understanding of complex infrastructure behaviour, extending as a consequence, the
lifecycle and durability of networks and reducing the environmental impact. Furthermore, the validation of the
integration and usability of this tool into existing commercial platform (RAMSYS by MerMec,
http://www.mermecgroup.com/51/1/ramsys.php) and pilot tests will ensure the linkage to available tools in order to
meet all the requirements of the rail sector.
 Manuel Menéndez et al. / Transportation Research Procedia 14 (2016) 1987 – 1995 1989

OPTIRAIL contributes to obtaining higher levels of safety and service in railway infrastructures, “optimal” life
cycle for the management of railway infrastructure maintenance, better quality of service and, therefore, higher level
of client satisfaction, improved level of availability of the railway infrastructure.
OPTIRAIL also ensures a more effective planning of the management and activities of infrastructure
maintenance based on expert knowledge accumulated over years of experience and to the information stored in the
monitoring and maintenance management systems. The automation of the maintenance management and task work
flow will improve the technical and economic effectiveness of infrastructure maintenance and management
minimizing the environmental impact.

Ensure cross-border coordination among European railway corridors

The European Commission (EC) issued several directives aimed at removing technical barriers to the supply of
equipment and the running of trains between Member States in order to allow interoperability (or technical
compatibility) of infrastructure, rolling stock, signalling and other rail systems.
These Directives provide specific details regarding the proper implementation of the interoperability
requirements. They include the authorisation of placing in service, the EC checking procedure and the EC
declaration of conformity with the essential requirements and the TSIs, Technical Specifications for Interoperability,
as well as the definition of the role of the notified bodies and cooperation between them.
To address those policies, the OPTIRAIL project has analysed in depth specifications and regulatory issues
across Europe. The in-depth analysis of data from two corridors’ case studies (Spain & Sweden) will allow finding
synergies between railway networks within corridors, as well as defining standards and common best practices for
the coordinated maintenance of their respective infrastructures.
Moreover, the transferability of tools from other sectors facing similar interoperability problems allows
OPTIRAIL to know and estimate solutions focused on coordinated decision-making and, consequently, fostering
interoperability and the promotion of a single European Railway Network.
Finally, the implementation and validation of the developed tools on pilot tests along the selected case studies,
together with the dissemination of results and the constitution of an Asset Management Club ensures the further
deployment or replication in other non-participating countries.

2. Objectives

The general objective has been to develop new smart tools to support the decision making towards optimizing
railway maintenance operations. The overall benefits of this process are ensuring higher railway availability, better
cross-border coordination and improved efficiency. The tools are built out of the available knowledge and know-
-how of the domain.

The specific objectives are:

x To analyse the specification and requirements for an efficient, reliable and sustainable maintenance of the railway
systems (for both passengers and freight), paying special attention to corridors in the selected case studies, both at
the system level (technical and regulatory) and the end-users level (know-how and tools for infrastructure
managers and track work contractors).
x To analyse the transferability of concepts, ideas and solutions from other sectors (electricity, oil & gas, water and
airspace) to be applied to this railway domain.
x To design and validate smart and collaborative tools based on intelligent systems (and particularly, on Fuzzy
Logic) that become decision support and making systems for railway maintenance.
x To design a new maintenance conceptual framework based on fuzzy systems that consider every functional and
technological component to ensure the integration and linkage to existing maintenance platforms and ICT
systems, and its performance and usability for infrastructure managers and track work contractors.
x To validate the design of the maintenance framework through an implementation and integration, total or partial,
with an existing maintenance platform.
1990 Manuel Menéndez et al. / Transportation Research Procedia 14 (2016) 1987 – 1995

x To demonstrate the usability of the project outcomes on existing rail networks (case studies).
x To define the corresponding models for facilitating the broad deployment of the project results.
x To define the synergies and best practices between the cross-border networks and disseminate the project results
to other corridors.

The operational objectives to ensure the objectives defined above are the following:
x Generation of bases of knowledge on regulations, policies, etc., regarding the railway maintenance. ICT
implementation of these bases of knowledge to be used by the maintenance platform.
x Characterization of the main maintenance tasks and their impact from different points of view: economic, human
and technical resources, serviceability, safety, environmental impact, etc.
x Development of models of behaviour and evolution of individual components and the whole infrastructure. This
will provide the ability to prevent the appearance of security and service events that affect the asset by its own
process of evolution and/or aging.
x Development of models of detection and classification of events that affect the infrastructure, and obtaining of
a linguistic description of the domain in technical terms.
x Development of computational models of maintenance tasks implemented on the basis of expert knowledge.
x Development of computational models of decision making for railway maintenance.
x Development of multi-criteria optimal planning models (level of service, availability, economy, CO 2 footprint,
traffic, etc.) for maintenance.

3. Project progress beyond the state of the art

According to the methodologies that are used in the decision support systems (DSS) currently used in the
domains, these are based on procedures and steps for applying maintenance strategies. Unfortunately, no study could
be found on developing a decision support system for railway infrastructure maintenance from an integral and global
point of view, which is a prerequisite for many corridors to ensure optimal maintenance. To that end, the OPTIRAIL
project pursued the development of a comprehensive tool to manage all the elements relevant for track maintenance.
This tool would ensure the linkage to available tools in order to meet all the requirements of the rail sector.
OPTIRAIL will develop a smart framework for maintenance management to facilitate making effective and
efficient decision in maintenance of railway networks within corridors and will act as an essential component of
a decision support system for the task. This tool will combine a top-down approach, from standards and regulations
to interventions planning on the railways, with a bottom-up approach, from interventions to be carried out of
standards and regulations.
Railway inspection is one area where techniques and approaches connected with, and based on, Artificial
Intelligence (AI), Soft Computing (SC) and Computational Intelligence (CI) have been deployed. These techniques
take advantage of collected data and the expert knowledge available to improve the timely effectiveness of
maintenance tasks. These techniques play an important role due to its performance, fast response and reliability to
support and enhance, from different points of view, the intervention of a human operator and consequent
performance. Most of these techniques can be hybridised to optimize the joined performance and used so far in the
railway inspection and decision making domain:

x Data Mining: data mining comprises the extraction and interpretation of knowledge and information from large
sets of data. This permits to estimate and evaluate the business intelligence, in this case regarding the
maintenance and management of railways.
x Fuzzy Logic: fuzzy logic has become a successful approach to address complex problems. Fuzzy logic succeeds
in formulating human knowledge and experience, and using them for decision making. This allows managing and
representing imprecise information and vagueness that are usual in real-world contexts such as the railway
domain.
x Artificial Neural Networks: artificial neural networks are human brain based mathematical models. That offers
computers capabilities such as learning to be applied to pattern recognition, modelling processes, clustering, etc.
used in solving complex tasks.
 Manuel Menéndez et al. / Transportation Research Procedia 14 (2016) 1987 – 1995 1991

x Evolutionary algorithms multi-objective algorithms to optimize tasks of processes subject to possibly

contradictory multi-criteria or multi-objective.
x Statistical Learning Theory-based models, just like Support Vector Machines.
x Linear and non-linear models for time series modelling and forecasting: ARIMA, threshold Autoregressive
Models, Garch Models.

The OPTIRAIL project applies these techniques to design new tools that will help to optimize railway
maintenance operations ensuring higher railway level of serviceability and better cross-border coordination. The
concept of OPTIRAIL can be summarized on Figure 1: all the available knowledge about the domain is translated
into a computer based version:

Fig. 1. Optirail general concept.

4. Work plan

The OPTIRAIL project work plan has been conceived according to the Work Packages structure shown in the
Figure 2.

Fig. 2. Work plan diagram.

WP1: In-depth analysis of maintenance requirements for a smart maintenance system

In depth study about how the different administrations are currently performing the maintenance, Adif (1994),on
their railway tracks and how they have been changing and updating over time. All this information is the starting
point from where the OPTIRAIL tool should begin.
1992 Manuel Menéndez et al. / Transportation Research Procedia 14 (2016) 1987 – 1995

In addition, knowledge about what construction techniques are being used, different characteristics of the
elements, Adif (1991), Adif (1987) and Adif (1985), that have being fitted, or what standards governing the
maintenance of the tracks are necessary in order to define a common maintenance criteria for European railway
corridors.
The Computational Intelligence is based on the data so the consortium needed to know what different kind of
surveying has been done over the tracks, how they have been done and where they have been stored. All of this
combined with expert knowledge will let us use and apply intelligent techniques to predict a maintenance operation.

WP2: Analysis of the transferability of tools

In Work Package 2, we studied maintenance practices and particularly use of multi-objective maintenance in four
case industries. Electricity transmission & distribution, off-shore gas transmission, water distribution and aerospace
were selected as they are characterized by high-criticality and distributed infrastructures where maintenance actions
may be disruptive to users and costly for the providers.
We incorporated relevant best practices from the case studies in a smart maintenance framework and
maintenance experts assessed relevance and ease of implementation for the railway sector.
Finally, we considered technical aspects and implementability of tools, methods and formulations from the case
industries for possible adoption in the OPTIRAIL project.

WP3: Tool conceptual design

Nowadays, railway maintenance involves many different tasks based on the information and knowledge collected
from a wide range of heterogeneous sources, affecting the way each railway contractor or administrator deals with
maintenance. The issue regarding to the railway maintenance is based on three layers from the point of view of
OPTIRAIL:

x Physical layer
o Physical parts of architecture: sensors + railway sections
x Logical/data layer
o Include data warehouse and acquisition/processing software
x Business/domain logic: OPTIRAIL Layer (see Figure 3)
o Information is processed in order to reach project goals
o SOA (Services Oriented Architecture) based client-server architecture

Three level architecture:

x Level A: Data & Knowledge acquisition from ICT & Data warehouse
x Level B: Business logic, implemented based knowledge models
x Level C: User Interface, devoted to interaction Human-Machine

This conception is an open conception permitting new cases, models, monitoring systems, etc.
To acquire this diversity, the design of OPTIRAIL framework is based on an Open concept of Services Oriented
Architecture (SOA) and is defined and implemented to reach the final goal: support railway maintenance decision
making based on knowledge. This proposed framework is arranged on the three mentioned levels or layers:
Physical, Logical and Business layers, embracing from data collection, processing and modelling, to decision
making and interface with users.
Several quality qualifiers (such as safety index, degradation index and OPTIRAIL index) have been considered to
supervise the track conditions along cross-border railways corridors. Their implementation can aid to make
a simpler, easier and more collaborative domain in which railways corridors are managed and categorized by
common standards and qualifiers, allowing stakeholders to develop their business not only in their local and national
markets but international ones, improving the capabilities of European companies to interact in global markets.
 Manuel Menéndez et al. / Transportation Research Procedia 14 (2016) 1987 – 1995 1993

Fig. 3. Optirail system.

WP4: Development of Fuzzy and Computational Intelligence based models for maintenance management
Several data and knowledge based models have been developed in OPTIRAIL, based on different CI approaches
such as: Fuzzy rule-based systems, neural networks, random forests, support vector machines, evolutionary
algorithms, and other methods, to enable a better maintenance tasks and planning C. Bergmeir et al. (2013).
On the one hand, deterioration models together with maintenance operation effect models make it possible to
model the behaviour of the track and to predict future quality measurements, across different scenarios of possible
intervention.
On the another hand, a multi-objective optimization framework then allows us to use this knowledge to find
maintenance plans optimized for specified objectives, such as minimizing intervention cost and maximizing
capacity, at the same time, while guaranteeing safety.

Fig. 4. Optirail outputs.

1994 Manuel Menéndez et al. / Transportation Research Procedia 14 (2016) 1987 – 1995

WP5: Integration and Usability validation of models

In order to respond to the needs of the different demonstrators and applications, a framework was created with
a high degree of adaptability and flexibility.
The OPTIRAIL tool is a decision support system whose core components are the intelligent modules developed
along WP4. Thereby, it enables the infrastructure manager to make maintenance decisions with improved
knowledge, and to assess different scenarios and assumptions.
A service oriented architecture with modular design and open source software gives the possibility to create
a community and to stimulate the market through new products and services based on OPTIRAIL.

Fig. 5. OPTIRAIL tool.

WP6: Case studies’ pilot tests

Once the proposed solutions have been integrated and tested on a commercial platform, WP6 (Pilot tests) will
consist of final testing stage of the integrated system. Two pilot tests were carried out in Sweden (Trafikverket) and
Spain (Adif) in order to validate the usability of the developed tools in the railway infrastructure D. Galar et al.
(2013).
The main approach for the validation is a variant of Living Lab (end-user involvement in systems design and
development) specifically adapted for validation and evaluation of rail industry scenarios and functionality demands.
By involving the end-user actively in the validation and tools design and development the results will be more
adapted to the “real-world” usage of the resulting tools and systems. Thus, the main objective for this work package
is to support the development of as usable tools as possible for the end-users and involved organizations.

5. Conclusions

Rail transportation is a key active within the European Union. To ensure and enhance its usability, an effective
maintenance is necessary. Automation and right decision making are essential to boost railway infrastructure
maintenance.
With the goal of developing an integrative smart framework to support railway maintenance the OPTRAIL
project was conceived. With the funding from EU, this project has been carried out, reaching its stated goals.
 Manuel Menéndez et al. / Transportation Research Procedia 14 (2016) 1987 – 1995 1995

Thanks to the developments carried out in OPTIRAIL a new maintenance model, based on Computational
Intelligence techniques, have been created. This system builds models for railway infrastructure and elements upon
which the deterioration of the assets is predictive. With a knowledge base extracted from experts, and the
deterioration models an optimal maintenance schedule is produced.
The accuracy of the derived models has a clear dependency on the quality and availability of data. The overall
system and methodology of OPTIRAIL has been validated. The effectivity of the systems for the case studies can be
greatly enhanced through the use of larger datasets. This is a key requirement to have a good models and thus make
effective maintenance decisions. In addition, the open and adaptive architecture of the OPTIRAIL framework allows
it for performance improvement along the time, with a continued learning process.
In order to reach a single European market, common practices should be defined, as a common interpretation.
There is a need of use the same language along the corridors, and artificial intelligence could be the key to develop
standard practices and procedures.

Fig. 6. Sweden and Spain pilot test railways.

Acknowledgements

This paper is derived from work carried out in the OPTIRAIL project, that has received funding from the
European Union’s Seventh Programme (FP7) for research, technological development and demonstration under
grant agreement No. 314031.

OPTIRAIL web site: http://www.optirail.eu/optirail/

References

Adif, NRV: 7-5-0.1. Conservación de la vía, criterios básicos sobre el mantenimiento de la vía, 1994.
Adif, N.R.V. 3-6-0.0. Descripción general- La vía y su material, 1991-Desvíos, N.R.V. 3-4-0.0. La vía y su material- Balasto y Subbase –
Balasto – Características determinativas de la calidad, 1987, N.R.V. 7-3-0.0. Trabajos en la vía- Calificación de la vía – Geometría de la vía,
1985.
Bergmeir, C., Sainz, G.I., Martínez, C., Benítez, J.M. “A Study on the Use of Machine Learning Methods for Incidence Prediction in High-Speed
Train Tracks”, Proc. Onf th 26th Intern. Conf. Industrial Engineering and Other Applictions of Applied Intelligent Systems,(IEA-AIE 2013),
pp. 674–683, Amsterdam, June 2013.
Galar, D., Kumar, U., Villarejo, R. and Johansson, C.A. “Hybrid prognosis for railway health assessment: An information fusion approach for
PHM deployment”, in Prognostics and System Health Management Conference (PHM 2013), vol 33 of Chemical Engineering Transactions,
pp. 769–774, Milan, September 2013.
Klir, G., Yuan, B. “Fuzzy Sets and Fuzzy Logic: Theory and Applications”, Prencie-Hall Press, 1995.
Kruse R. and Borgelt, C. “Computational Intelligence: A Methodological Introduction”, Springer, 2013.
Pace P. and Kontokostas, L. “Railway infraestructure asset and condition data collection, data usage and deterioration modelling for, maintenance
performance measurement”, in 10th World Congress on Railway research (WCRR 2013), pp. 1–6, Sydney, November 2013.