RAILWAY MODELLING

The Case for Ontologies in the Rail Industry

J. M. Easton, J. R. Davies and C. Roberts

School of Electronic, Electrical and Computer Engineering, University of Birmingham, Birmingham, U.K.

Keywords:

Railways, Data exchange, Domain ontology.

Abstract:

As the demand for rail travel grows amongst the travelling public, capacity on the largely Victorian infrastruc-

ture is becoming an important issue. High construction costs and the need for environmental responsibility

mean that simply building more lines is not necessarily a viable option, and therefore the rail industry is be-

ginning to look for ways in which their current assets can be used more intelligently. In the ﬁrst wave of this

process, a number of new projects have been started that monitor the condition of vehicles and infrastructure,

but the greater challenge of integrating that data and using it to support business operations is still largely

unaddressed. This paper outlines a number of data handling and data sharing issues that face the rail industry,

and presents the argument for the adoption of an ontology-based data standard across the sector.

1 INTRODUCTION

As rail transport becomes an increasingly attractive

option to the travelling public, offering a fast, reli-

able means to beat trafﬁc jams and travel straight into

major urban centres, issues of capacity on the largely

Victorian railway network in Great Britain are of ever

greater importance. A recent report has shown a 7.3%

increase in passenger kilometres between 2007-08

and 2008-09 (Ofﬁce of Rail Regulation, 2009). The

huge costs associated with alterations to infrastructure

mean that increasing capacity through the large-scale

construction of new lines is not always a viable op-

tion, and as a result the railway industry acrossEurope

is beginning to look at ways in which the railways can

be made smarter.

As part of the move towards a smarter railway,

Network Rail (the owner of Britain’s rail infrastruc-

ture) has started two complementary new initiatives

- Intelligent Infrastructure (IIS) and LiveTrain. The

Intelligent Infrastructure project, currently being tri-

aled on a stretch of line between Edinburgh and Glas-

gow, monitors the condition of ﬁxed assets such as

track circuits, points, and signalling systems, pro-

cessing the data and raising alarms for operators via

a supervisory control and data acquisition (SCADA)

system. The LiveTrain project by comparison, will

see hundreds of older passenger and freight trains on

the British railway network retroﬁtted for condition,

power and location monitoring. In the ﬁrst instance

it is hoped that these projects will improve the re-

liability of existing assets, leading to more efﬁcient

maintenance practices and reduced delays across the

network.

2 ISSUES AFFECTING DATA

SHARING IN THE RAIL

SECTOR

The data that will be collected by projects such as

IIS and LiveTrain is an important step towards im-

proving performance on the smarter railway; how-

ever, the greater tasks of storing, processing and pre-

senting the information remain to be addressed. In

the British railway industry these tasks are made more

complex by a number of key factors, includinga range

of legacy systems, a large number of competing stake-

holders, and historical working practices. When look-

ing further aﬁeld to the European, US and Asian rail-

way networks, the situation is further complicated

by differences in nomenclature and units of measure-

ment. In order to properly put forward a case for an

ontology-based data standard in the sector, we must

ﬁrst elaborate on these issues.

257

M. Easton J., R. Davies J. and Roberts C..

RAILWAY MODELLING - The Case for Ontologies in the Rail Industry.

DOI: 10.5220/0003092102570262

In Proceedings of the International Conference on Knowledge Engineering and Ontology Development (KEOD-2010), pages 257-262

ISBN: 978-989-8425-29-4

 2010 SCITEPRESS (Science and Technology Publications, Lda.)

2.1 Legacy Systems

Data logging in general, and the monitoring of the

condition of assets in particular, is not a new idea on

the British railway network (although IIS and Live-

Train will perform these tasks on a much larger scale

than was previously the case). Both the infrastruc-

ture operator, and various train operating companies

have some ability to monitor the condition of their as-

sets. Network Rail for example, has the WheelChex

system, which records the impact force that a train’s

wheel makes on specially instrumented sections of

track, and hot axle box detectors, which can moni-

tor the temperature of the axles on passing vehicles in

order to identify potential problems. These systems

provide useful information and represent a signiﬁcant

amount of previous investment. Certainly, for man-

agement reasons if nothing else they must form an

important part of the smarter railway. Unfortunately,

many of the current monitoring systems were devel-

oped and installed in isolation, and may also be oper-

ated by third-parties. The raw data, if available at all,

is in proprietary formats, greatly increasing the com-

plexity of any attempt to combine it with data from

other sources.

Equipment longevity on the rail network is ex-

pected to be high. In 2007 the average age of rolling

stock on the British network was around 13 years,

one of the lowest ﬁgures in Europe (Department for

Transport, 2007). On vehicles that already have con-

dition monitoring capabilities, different manufactur-

ers may have chosen to instrument different systems,

meaning that there is no standard set of measurements

that can be used as a common basis for metrics across

the network. Trackside equipment can be much older,

with some side lines still using semaphore signalling

instead of the newer electronic signals.

Similar problems can be seen with software sys-

tems, a good example being the Total Operations Pro-

cessing System (TOPS) used by the Train Operat-

ing Companies to store information on locomotives.

TOPS was created in the 1960s to run on mainframe

computers, is written in its own programming lan-

guage, and generates rather cryptic plain-text reports.

2.2 Stakeholders

The privatisation of the British railway network in the

late 1990s has led to a confusing array of different

stakeholders being responsible for different elements

of the business. Amongst these entities are statu-

tory authorities including the Department for Trans-

port and Ofﬁce of Rail Regulation, standards bodies

such as the Rail Safety and Standards Board (RSSB),

the train operating companies, rolling stock leasing

companies, infrastructure operator, passenger groups,

unions, and countless contracted maintenance and

construction companies. Each stakeholder group has

different interests in the network, operating practices,

and levels of summary at which they need to dis-

play various items of data. Some of them, such as

the train operating companies, may be in competition

with each other and be unwilling to share informa-

tion. Although the British system is more complex

than most, railways throughout Europe (and the rest

of the world) are tending to follow this trend.

2.3 Historical Practices

The safety critical approach which must be taken

when dealing with any new addition to the railway

means that there are numerous practices, procedures,

data gathering methods, and data storage systems that

must be maintained. Many of these were created be-

fore computing became main stream and present dif-

ﬁculties, both technical and ﬁnancial, that must be

overcome to create a uniﬁed data transfer mechanism.

Particular challenges are posed by records that are

only kept on paper, and information exchanges be-

tween stakeholders that are performed over the tele-

phone. Currently, this information is essentially lost

and can not be used in the future to inform business

decisions.

2.4 Nomenclature

While the use of terms within the railway industry in

a particular country are usually consistent, this is not

necessarily the case when considering railways in dif-

ferent parts of the world. The terminology used in the

British and US railways for example have a number of

major differences. In Britain a shunt is an operation

when coaches or trucks are moved from one track to

another, usually to change the formation of a train; the

equivalent US term for shunting is switching. While

ambiguity of this type can usually be overcome by hu-

mans, it can be a very signiﬁcant problem when au-

tomatically exchanging data between computer sys-

tems; an XML ﬁle with a tag “shunt” generated in

Britain for example, may not have the same meaning

as the tag in a ﬁle that originated in the US.

Although not strictly an issue on nomenclature, an

important point when considering the difference be-

tween the British and European railway networks lies

in units of measure. While European railways tend

to be measured in metric units, the British system

is measured according to imperial miles and chains

for infrastructure purposes. Uncommunicated differ-

KEOD 2010 - International Conference on Knowledge Engineering and Ontology Development

258

ences in units of measurement have been known to

have catastrophic effects in the past, for example in

the loss of the Mars Climate Orbiter (Mars Climate

Orbiter Mishap Investigation Board, 1999).

3 ONTOLOGIES AS A DATA

STANDARD FOR THE RAIL

INDUSTRY

While it is clear that something must be done to im-

prove data exchange and sharing within the industry

in order to bring about the smarter railway, what that

action should be may be less clear. It is unlikely that

any one solution will be a “magic bullet”, resolving

all the factors outlined above, but what are the op-

tions? Ruling out the replacement of current systems

on a large scale, mainly on the basis of cost and the

rigorous requirements for ensuring safety within the

sector, we are left with creating a data transfer mech-

anism that is capable of meeting the needs of all the

current solutions. From a technical perspective, on-

tologies have a number of advantages to offer: data

transferred according to an ontology retains its con-

text, making it possible for machines to reason on it

and for intelligent software agents to perform a wide

range of data processing tasks autonomously. Ontolo-

gies are also easily extended by individual companies

to meet their own needs, an important consideration

when they are in competition with each other and may

need to protect elements of intellectual property. Fi-

nally, software that communicates using an ontology

can interact with other packages as part of the seman-

tic web.

3.1 Business Case

In the British railway industry, developing a solid

business case for a major project is second only in im-

portance to safety concerns. The following sections

will discuss the potential advantages of an ontology-

based data exchange standard to the industry.

3.1.1 Condition Monitoring and Predictive

Maintenance

The beneﬁts of the use of ontologies with regard to

condition monitoring and predictive maintenance in

railway vehicles has already been demonstrated in

work done as part of the InteGRail project (Lewis

and Roberts, 2010). Currently, maintenance in the

British railway industry is performed according to a

schedule; vehicles are taken to a depot and exam-

ined at speciﬁed intervals, irrespective of whether

there is an apparent problem at the time. A lack of

condition monitoring data means that failures, when

they happen, are dealt with using a ﬁreﬁghting ap-

proach. As part of the LiveTrain project, much more

condition monitoring data on older vehicles will be-

come available. By transferring condition monitor-

ing data according to an ontology model, ontological

inference can be used to generate a consistent set of

data items from vehicles with differing instrumenta-

tion sets. From this position it is much easier to gen-

erate metrics for the prediction of faults, since more

data of a consistent form is available to train the clas-

siﬁer being produced. If it can be shown that the clas-

siﬁcation algorithm can predict faults a sufﬁciently

long period before failure that maintenance work can

be performed, then the scheduled maintenance pro-

cesses, which are costly and potentially unnecessary,

can be safely reduced in frequency.

The ability to predict the failure of vehicles should

also lead to a reduction in the number of in-service

failures. On the British railway network train oper-

ating companies are charged by the minute for late-

running services; as approximately 20% of delay min-

utes for the year 2006-07 were attributed to ﬂeet

causes (Department for Transport, 2007), a reduc-

tion in breakdowns has important ﬁnancial implica-

tions for the train operators. These are aside from the

customer service, timetabling and logistics difﬁculties

that arise from a train being out of position.

3.1.2 Decision Support

While the large-scale gathering of information by

projects such as IIS and LiveTrain is an important

step towards the smarter railway, its value is never re-

alised if it is not used by an organisation to effectively

inform its business decisions. This is a complex is-

sue with many different facets that must be consid-

ered; individuals with different roles within the in-

dustry may, for example, require radically different

views of the information available to perform their as-

signed tasks. Data volume is also an important issue,

and the information must be ﬁltered and summarised

so that humans, who are ultimately responsible for

the decision-making process, are not overloaded with

information that is not relevant. In a system where

data is transferred according to an ontology model,

the context of the data is retained. As a result of this,

the combination of elements of the data for particu-

lar tasks could ultimately be performed by software

agents, making the software applications themselves

simpler and more easily extended and maintained.

RAILWAY MODELLING - The Case for Ontologies in the Rail Industry

259

3.2 Interoperability

While much of the current European rail network has

evolvedas a set of national systems, the realities of the

European Union have meant that travelling across Eu-

rope is now much easier. With that in mind, the Euro-

pean Parliament has, for the last 10 years, been pass-

ing directives designed to encourage interoperability

between the nation rail networks (European Parlia-

ment, 2001). The issues of nomenclature, units of

measurement etc. outlined above make an ontology-

based data transfer standard essential here, as the

exact nature of the data is recorded in a machine-

interpretable format, allowing appropriate steps to be

taken in software to ensure that data is presented and

processed using the correct scales.

3.2.1 Capacity Planning and Timetables

Railway scheduling on a national network is a com-

plex task; among the factors that must be considered

are demand for a service, peak travel times, track and

vehicle maintenance, speed limits, load limits, con-

necting services, timings of other services on the same

line, and overall journey times. The situation is even

more involved when considering cross border routes,

here the individual responsible for scheduling the ser-

vice must have information on the railway network in

each country the proposed service will pass through.

Changes to local services in a country could easily

have knock-on effects on services passing through it,

although the way services and infrastructure are de-

scribed will almost certainly change in the various na-

tional computer systems; this presents a clear need for

a common mechanism for sharing route and schedul-

ing information in a timely and unambiguous manner.

3.3 Customer Experience

The potential effects of an ontology-based data trans-

fer standard for the rail industry on the customer ex-

perience can not be understated. While many of the

business case beneﬁts outlined above do, of course,

impact on the customer experience (interoperability

between networks, reduced delay minutes, and im-

proved reliability), the most marked improvements

will come through the rise of the semantic web. As

part of the smarter railway, customers wishing to book

train tickets to travel to a meeting for example, will

only need to inform a software agent of their intended

destination, and it will be able to ﬁnd tickets for them.

This might involvenegotiating with the customer’sdi-

ary and interpreting the timetable to ﬁnd the best time

for their journey, arranging for overnight accommo-

dation in their preferred hotel, and booking a table for

dinner. As the customer travels, agents will keep track

of their progress and be able to adjust their itinerary

if they are delayed or miss a connection, updating the

time of the meeting if that ﬁts with the diaries of the

other parties attending. Obviously the smarter railway

is only a small part of the semantic web vision, but it

is an important element nonetheless.

4 THE CREATION OF AN

ONTOLOGY TO MODEL THE

RAIL INDUSTRY

When considering a modelling task of this scale, there

are a number of important factors to be addressed.

These include practical issues such as the sources of

information available and compatibility with existing

initiatives, along with the more strategic questions

of how to get enough support from users to achieve

a critical mass and force large-scale adoption of the

model. The remainder of this paper discusses some

initial thoughts on these matters, before going on to

identify the key areas that should be captured in an

initial model.

4.1 Industrial Support

As any data transfer standard will require buy-in from

large proportions of the industry in order to be suc-

cessful, it is vital that it reﬂects the needs of as large a

group of the rail stakeholders as possible. To that end,

it is suggested that a consortium be formed, consisting

of members of rail governing bodies, operating com-

panies, developers of existing standards and industrial

partners to help guide the creation of the ﬁnal model.

In order to promote a sense of ownership within the

industry, as much of the development work as pos-

sible should follow a collaborative approach, poten-

tially even to the extent of using a community-based

ontology engineering methodology like DILIGENT

(Pinto et al., 2004b; Pinto et al., 2004a; Tempich et al.,

2007). For collaborativemodelling efforts to be effec-

tive however, it is desirable to have an initial “straw

man” that can be used to guide the subsequent dis-

cussions. As such, in advance of the formation of a

consortium, work has begun in Birmingham to inves-

tigate the available sources of domain knowledge.

4.2 Sources of Domain Knowledge

A number of sources of domain knowledge, aside

from the expertise of the consortium members, are al-

ready in existence within the rail industry that could

KEOD 2010 - International Conference on Knowledge Engineering and Ontology Development

260

be drawn on to speed the creation of an initial model.

These include, but are not limited to, existing ontol-

ogy models, XML data standards, technical reports,

and legislative guidelines.

Existing Ontologies

The InteGRail project (Fischer et al., 2009; Lewis

and Roberts, 2010) was a European FP6 programme

with a brief of “integrating current and future railway

information systems across the European nations in

an effort to achieve improved overall efﬁciency and

performance, particularly in future European railway

endeavours”. As part of the project, a basic condi-

tion monitoring ontology for wheel impact faults was

created, which could reason about the severity of re-

ported faults and classify a vehicle status accordingly.

The ISO 15926 standard (Batres et al., 2007) was

initially developed for the integration and exchange

of information relating to process plants including oil

and gas production facilities. ISO 15926 takes a very

“ground-up” approach to the modelling of processes,

allowing each piece of equipment to be described in

terms of its component parts as well as its tempo-

ral existence. Recently, an Ontology Web Language

(OWL) implementation of the ISO 15926 standard

has been produced, which could serve as an upper-

level ontology for the project.

Data Exchange Standards

The RailML (Nash et al., 2004) initiative provides

an XML based approach to transmitting and receiv-

ing rail related data. The proposed standard covers a

range of rail terminology, including vehicles, subsys-

tems and infrastructure, and provides not only a good

source of domain concepts and instances, but also ba-

sic hierarchies.

The Maintenance Information Open System Al-

liance (MIMOSA) are a trade association “dedicated

to developing and encouraging the adoption of open

information standards for operations and maintenance

in manufacturing, ﬂeet, and facility environments”.

Among the family of standards they produce is the

Open System Architecture for Condition Based Main-

tenance (OSA-CBM) (Lebold and Byington, 2002),

an ISO 13374 compliant data transfer architecture al-

ready being considered for use within the British rail-

way industry. While the MIMOSA standards use their

own data model, and may therefore not be suitable for

direct use within the project, they do provide a good

model for the implementation of an ISO 13374 com-

plaint system.

Standards Bodies

In Britain there are several organisations that issue

standards (both general and rail speciﬁc). The RSSB

publish a set of safety related standards collectively

referred to as the Railway Group Standards (RGS).

These include approved codes of practice, guidance

notes and railway industry standards. The British

Standards institute (BSi) also issue a broad range of

railway-applicable standards, some speciﬁc to the rail

industry, others relating to broader topics. Many of

these standards are closely related to standards issued

by the International Organization for Standards (ISO).

RailML ISO 15926

MIMOSA

OSA-CBM

InteGRail

Ontology for semantic

interoperability in the rail

industry

Industrial

Partners

Upper-level

Ontologies

Figure 1: Sources of knowledge for a rail domain ontology.

4.3 Priorities for an Initial Model

The scale of the modelling task proposed by this pa-

per should not be underestimated, and it is inevitable

that certain areas of the domain would need to be pri-

oritised during the development of any ontology. For-

tunately, within the rail industry a clear group of top-

ics exist that not only represent the minimum require-

ments for a useful data transfer standard, but would

also facilitate a signiﬁcant proportion of the business

beneﬁts discussed earlier; those being vehicles (both

as individuals and as train formations), infrastruc-

ture (track types, operating speeds, clearances, power

supply and signalling) and timetabling (including the

scheduling of services).

Although these topics still represent a substantial

implementation effort, they are all required in order

to describe many events on the railway network. As

an example consider the activation of trackside mon-

itoring equipment such as a WheelChex device. The

WheelChex system itself is part of the infrastructure,

consisting of a section of instrumented track at a par-

ticular location. The defective wheel which activated

WheelChex would be part of a particular train for-

mation, but in order to uniquely identify it reference

would have to be made to both the timetable for the

line and to the details of the train formation so the

individual vehicle (coach, wagon, locomotive etc.)

could be singled out. Once the vehicle had been iden-

tiﬁed, the tracks which it had run over since it last suc-

RAILWAY MODELLING - The Case for Ontologies in the Rail Industry

261

cessfully passed a WheelChex device (during which

it may have been part of a number of different train

formations) would need to be determined so that they

could be checked for damage, again requiring access

to the timetable, records of train formations, and the

details of the tracks.

As the example shows, information on vehicles,

infrastructure and timetabling are sufﬁcient to report

the activation of lineside equipment. Information

from the same topics would also allow the scheduling

of services to be performed, where the vehicles pro-

posed for use must be checked for compatibility with

the infrastructure over which they are to run (clear-

ances to lineside equipment and other vehicles, track

gauge, availability of power), and an appropriate free

timeslot must be identiﬁed in the timetable. It would

also allow basic semantic web services to be created,

improving the customer experience.

On completion of the initial model the ﬁrst phase

of expansion should include information on mainte-

nance practices, allowing the correct responses to the

activation of WheelChex and similar equipment to be

determined, and details of the various roles that em-

ployees have within the industry, enabling appropri-

ate information to be displayed for each worker. This

could be followed later by business logic, which al-

though providing the key ﬁnancial beneﬁts to the in-

dustry, would serve little purpose without the data

modelled by the earlier phases.

5 CONCLUSIONS

An ontology-based standard for data transfer would

have much to offer the rail industry, allowing existing,

largely incompatible, legacy systems to exchange in-

formation in a meaningful way, without the need for

costly changes that would potentially pose a risk to

safety. More importantly, it would provide a frame-

work on which modern, semi-autonomous processing

agents could be built, improving the efﬁciency of the

railway network, reducing the risk of human errors in

mundane tasks, and enhancing the experience of the

travelling public. This paper has attempted to present

the arguments for, and the challenges to be addressed

during, the creation of such a standard; along with

some initial thoughts on the methods that could be

adopted for its development.

REFERENCES

Batres, R., West, M., Leal, D., Price, D., Masaki, K., Shi-

mada, Y., Fuchino, T., and Naka, Y. (2007). An Upper

Ontology Based on ISO 15926. Computers & Chemi-

cal Engineering, 31:519–534.

Department for Transport (2007). Delivering a sustainable

railway. Technical report, Department for Transport,

http://www.dft.gov.uk/about/strategy/whitepapers/whi

tepapercm7176/.

European Parliament (2001). Directive 2001/16/EC of the

European Parliament and of the Council of 19 March

2001 on the interoperability of the conventional

rail system. Technical report, European Parliament,

http://europa.eu/legislation summaries/internal marke

t/single market for goods/technical harmonisation/l2

4229 en.htm.

Fischer, J., Roshchin, M., Langer, G., and Pirker, M. (2009).

Semantic data integration and monitoring in the rail-

way domain. In Proceedings of the 10th IEEE interna-

tional conference on Information Reuse & Integration,

pages 11–16.

Lebold, M. and Byington, C. (2002). OSA-CBM architec-

ture development with emphasis on XML implemen-

tations. In Proceedings of the 2002 Maintenance and

Reliability Conference (MARCON).

Lewis, R. and Roberts, C. (2010). Using non-monotonic

reasoning to manage uncertainty in railway asset

diagnostics. Expert Systems with Applications,

37(5):3616–3623.

Mars Climate Orbiter Mishap Investigation Board

(1999). Mars Climate Orbiter Mishap Investiga-

tion Board Phase 1 Report. Technical report, NASA,

ftp://ftp.hq.nasa.gov/pub/pao/reports/1999/MCO repo

rt.pdf.

Nash, A., Huerlimann, D., Schuette, J., and Krauss, V.

(2004). Computers in Railways IX, chapter RailML: a

standard data interface for railroad applications, pages

233–242. WIT Press.

Ofﬁce of Rail Regulation (2009). National Rail

Trends: 2008-2009 Quarter One. Technical re-

port, Ofﬁce of Rail Regulation, http://www.rail-

reg.gov.uk/upload/pdf/382.pdf.

Pinto, H., Staab, S., and Tempich, C. (2004a). DILIGENT:

Towards a ﬁne-grained methodology for DIstributed,

Loosely-controlled and evolvInG Engineering of oN-

Tologies. In Proceedings of the 16th European Con-

ference on Artiﬁcial Intelligence (ECAI 2004), pages

393–397.

Pinto, S., Staab, S., Sure, Y., and Tempich, C. (2004b). On-

ToEdit empowering SWAP: a case study in supporting

DIstributed, Loosely-controlled and evolvInG Engi-

neering of oNTologies (DILIGENT). In Proceedings

of the 1st European Semantic Web Symposium (ESWS

2004), pages 16–30.

Tempich, C., Studer, R., Simperl, E., Luczak, M., and Pinto,

H. (2007). Argumentation-based ontology engineer-

ing. 22(6):52–59.

KEOD 2010 - International Conference on Knowledge Engineering and Ontology Development

262