INTEGRATING RAILWAY MAINTENANCE DATA
Development of a Semantic Data Model to Support Condition Monitoring Data
from Multiple Sources
J. Tutcher, C. Roberts, and J. M. Easton
School of Electronic, Electrical and Computer Engineering, University of Birmingham, Birmingham, U.K.
Keywords:
Railways, Data exchange, Domain ontology, Data interoperability, Condition monitoring.
Abstract:
Railway networks comprise a large number of information systems, many of which are implemented by differ-
ent stakeholders according to different design requirements, and in different ways. Owing to the safety-critical
nature of these systems, data is rarely shared across boundaries, and the potential for re-use of information is
lost. Using ontology, it is hoped that information from these systems can be extracted and shared, in order to
facilitate better operational decision-making. This paper examines the aspects of data re-use likely to benefit
the industry, and describes a railway condition monitoring ontology that is being designed in conjunction with
several industrial stakeholders to improve operational efficiency.
1 INTRODUCTION
Railway networks around the world are of many dif-
ferent information systems, supplied and controlled
by different stakeholders. Owing to the complexity
of these systems, and the vastly different design ap-
proaches taken in each case, very little information
sharing occurs, leading to duplication of data and sub-
optimal decision-making in many areas. With railway
usage havingincreased steadily in most countriesover
the past decade (Office of Rail Regulations, 2010),
and with this trend set to continue, there is now a de-
mand for greater and more intelligent utilisation of
capacity and resources, and it is possible that much-
needed efficiency gains in operation and maintenance
of railways can be achieved through facilitating the
sharing and exchange of information across the net-
work.
This paper discusses the application of ontol-
ogy for integrating day-to-day railway control data
(specifically train locations and characteristics) with
railway condition monitoring (RCM) system data.
Various condition monitoring methods are in use
throughout the world, including a number of points
monitoring systems, signal lamp monitoring, and
axle monitoring (for overloaded or over temperature
axles). None of these, however, currently utilise the
wealth of operational data available through other sys-
tems in order to provide context to information pre-
sented, and it is up to human operators to spend time
researching associated train information should any
of these systems present a fault - a task which is slow,
error-prone, and costly. Through work undertaken
with Invensys Rail Group (IRG), a manufacturer of
railway signalling and control systems, it is the au-
thors’ aim to research and show that not only can rel-
evant railway operations data be automatically pre-
sented with RCM fault notifications, but also that a
common data model for various RCM systems can
be established in the form of an ontology. This will
be done through the development of a semantic data
model to integrate data from InvensysRail Group sig-
nalling and control systems with existing RCM sys-
tems currently in use on the railways.
1.1 Industrial Support
Invensys Rail Group (IRG) is a railway signalling
and control company, whose products and systems
are ubiquitous across railway networksthroughoutthe
world. This project is being undertaken with support
from IRG, allowing knowledge gained throughout to
be implemented in future products and systems. In
addition to IRG’s support for this project, Network
Rail in the UK are also exploring the capabilities of
more effective condition monitoring and data man-
agement in order to maintain the railway network
more effectively, most notably through their Intelli-
gent Infrastructure (IIS) project (Ollier, 2006).
442
Tutcher J., Roberts C. and M. Easton J..
INTEGRATING RAILWAY MAINTENANCE DATA - Development of a Semantic Data Model to Support Condition Monitoring Data from Multiple Sources.
DOI: 10.5220/0003661404420444
In Proceedings of the International Conference on Knowledge Engineering and Ontology Development (KEOD-2011), pages 442-444
ISBN: 978-989-8425-80-5
Copyright
c
2011 SCITEPRESS (Science and Technology Publications, Lda.)
2 EXISTING SYSTEMS
Multiple RCM systems are currently in use across the
UK rail network, to varying degrees of complexity. In
modern signalling installations some power and track
circuit condition monitoring capabilities are already
built into control software, but the extent to which
these are integrated with maintenance work is still
limited. Beyond a signallers’ warning to the train
driver, no alerts are issued by the system itself. The
types and capabilities of existing signalling and con-
trol systems vary widely across the UK rail network
and throughout the world; as a result, while one ben-
efit of an ontology-based system design is that multi-
ple system architectures can be dealt with easily, it is
likely that this project’s scope will be initially limited
to IRG’s WestCAD system.
As part of the InteGRail project (Langer et al.,
2008; Lewis et al., 2006), a European project aimed
at exploring integration between railways, a number
of ontology demonstrations were created showing the
capabilities of the technology in the rail domain. A
’core ontology was developed, encompassing high
level railway infrastructure information such as routes
and vehicle types, and several child ontologies de-
signed to explore particular use cases identified ini-
tially. One such ontology looked at the handling of
condition monitoring data from several existing data
sources, and used reasoning in order to establish ex-
plicit and implicit fault conditions on a particular area
of railway. Knowledge gained from the InteGRail ap-
plications will be built on in this project, specifically
in the case of the railway condition monitoring system
ontology design.
2.1 Wheel Impact Load Detection
Network Rail (owners and managers of UK mainline
rail infrastructure) have 26 wheel impact load detec-
tion (WILD) systems in use on their network. These
systems are placed at points along railway track to
assess the physical condition of trains as they pass,
in order to prevent further damage to railway infras-
tructure. Sensors measure stress/strain exerted on the
rails to determine information about a vehicle as it
passes. For each train passing a WILD site, date/time
of passage, track and direction, wheel loads, and axle
counts are recorded by the system and passed to a re-
mote PC (AEA Technology, 2008). Using the cur-
rent system, this information is passed to a central
computer, and then distributed to regional centres for
analysis using a piece of bespoke PC software. It is
up to human operators in each location to interpret
this data and decide which (if any) axles are dam-
aging or dangerous, as well as to identify the actual
trains passing the WILD system at any time ’ identifi-
cation is done based on known train movements using
TRUST, a Network Rail run computer mainframe-run
database (Network Rail, 2010) as well as through ed-
ucated guesses of train types from axle counts. This
is far from ideal, and the time taken to deduce train
location and fault conditions is considerably longer,
and less consistent, than should be possible using an
automated ontology-based alternative.
3 ONTOLOGY APPLICATION
3.1 The Case for Ontology
Projects such as InteGRail (Langer et al., 2008) have
recently shown applications of ontology across the
rail industry, and specifically in handling various
types of railway condition monitoring data (as in the
wheel impact load measurement / hot axle box detec-
tion ontology created as part of that project). The In-
teGRail project successfully showed the inference ca-
pabilities of ontology technology within the field; this
project aims in addition to explore data integration
between train control systems and condition moni-
toring. A correctly modelled ontology system also
has the more fundamental advantage of allowing data
collected to retain its meaning and context, allowing
more flexible querying and interrogation of data than
is currently possible.
3.2 Data Exchange & Interoperability
A major advantage of an ontology-based approach to
storing RCM data is that it can deal very well with
heterogeneous data. Even across multiple systems
designed by the same supplier, interpreting and act-
ing on fault data from different types of monitoring
equipment and sensors can be a challenge, and the
structure of an ontology-basedapproach lends itself to
coping with these scenarios very well. If the ontology
is constructed such that states from condition moni-
toring equipment can be deduced regardless of sys-
tem type (for instance inferring a ’fault’ condition in
each different type of system), communication is rel-
atively trivial. Across existing multi-vendor systems,
where data structures and communications protocols
are likely to differ more, a semantic model allows data
integrity to be maintained whilst building a common
platform on which to reason and observe. Ontology
also providesa flexible platform for expandingsystem
reach further into the system’s life, and will allow ex-
pansion into other areas of railway maintenance and
INTEGRATING RAILWAY MAINTENANCE DATA - Development of a Semantic Data Model to Support Condition
Monitoring Data from Multiple Sources
443
operations without the requirement for a drastic re-
structuring of the system.
3.3 Reasoning & Inference
Using an ontology model to combine condition mon-
itoring data with railway operations provides the po-
tential for autonomous reasoning, removing the ne-
cessity for human operators to spend large amounts
of time cross-referencing and checking raw data in
order to determine and describe vehicle/infrastructure
faults. This is likely to be a key benefit to railway op-
erators, and an aspect of the system which will be de-
veloped in depth. A reasoner will allow, for example,
the association between a vehicle’s geographic loca-
tion and an RCM fault value to be done automatically.
This is one of the key benefits of using an ontology-
based system to integrate data across the railway, and
knowledge inferred through reasoning techniques is
likely to be of great value to railway operators and
other stakeholders.
4 A WILD ONTOLOGY
Initially, the development of a new ontology system
to address the data integration issues discussed above
will aim to capture informationfrom the Network Rail
WILD axle load monitoring system and combine it
with train running data (such as vehicle locations and
descriptions). It is proposed that such an ontology
include knowledge of several aspects of the railway
domain, as follows:
Railway geography and network layout, in order
to locate vehicles on track and RCM sensors. In-
vensys Rail’s WestCAD system tracks train lo-
cations in discrete ’bays’ along the track, based
on data from railway track circuits or axel coun-
ters; the WestCAD data, or even track circuit data
could be used to locate trains in an ontology.
Railway vehicle knowledge. Train headcodes
contain limited information about vehicle class
(freight, passenger, high speed services); this in-
formation should be represented as such, with
the possibility of being expanded upon using data
available from other systems (in-cab diagnostics,
train operator data).
RCM knowledge such as WILD raw data, passing
vehicle type, fault conditions, equipment charac-
teristics, and equipment geography.
With these aspects correctly modelled, it is possible
that rules created in an ontology reasoner will allow
WILD data to be correlated to railway vehicles, and
could also associate vehicle types with common line
faults (for instance train bogie designs that are less
reliable than others). Train speed and weight as ac-
quired by the WILD system will also form part of
the ontology, allowing speed profiling of services over
long periods of time to be performed. Creation of this
data model will allow the retrieval of WILD infor-
mation with regards to specific trains, automatically,
and will thus save the expense of needing human re-
sources to verify all incoming RCM data. The even-
tual use of this system will add value to existing IRG
products, and continued development of the ontology
system should allow greater utilisation of the technol-
ogy in the railway domain.
4.1 Future Development
It is important that the WILD system as described
above be modelled in such a way that further develop-
ment is possible, in order that many more sub-systems
can be included and further knowledge inferred. It
is the project’s aim to use this ontology application
as a basis for a larger railway systems ontology over
the next two years, showing that further streamlining
of railway data can be achieved across control, op-
erations, and maintenance of the network. Further
work will involve the inclusion of other railway sys-
tems data in the ontology, such as recorded vehicle
’black box’ data (many trains record speed, location,
and diagnostics information as they travel), and com-
prehensive timetable information. With the addition
of these data sets, an ontology system will be able to
perform further reasoning in order to contribute new
knowledge to railway stakeholders.
REFERENCES
AEA Technology (2008). Wheelchex user guide.
Langer, G., Lewis, R., and Roberts, C. (2008). Ontology
driven railway rcm data integration. In Railway Con-
dition Monitoring, 2008 4th IET International Confer-
ence on.
Lewis, R., Fuchs, F., Pirker, M., Roberts, C., and Langer, G.
(2006). Using ontology to integrate railway condition
monitoring data. In Railway Condition Monitoring,
2006. The Institution of Engineering and Technology
International Conference on, pages 149 –155.
Network Rail (2010). Catalogue of railway code systems.
Office of Rail Regulations (2010). National rail trends
2009-10 yearbook.
Ollier, B. D. (2006). Intelligent infrastructure the business
challenge. In Railway Condition Monitoring, 2006.
The Institution of Engineering and Technology Inter-
national Conference on, pages 1 –6.
KEOD 2011 - International Conference on Knowledge Engineering and Ontology Development
444