EPSRC Studentship / Scholarship (Fully Funded)

A Computational Intelligence Based System for Track Maintenance

Department/School: Jointly between Department of Civil Engineering and School of Computer Science, the University of Birmingham, UK

Supervisors: Dr Michael Burrow (Civil Engineering) and Professor Xin Yao (Computer Science)

Contact: Dr Michael Burrow
Tel: (+44) (0)121 414 2626
E-mail: m.p.burrow@bham.ac.uk

Professor Xin Yao
Tel: +44 121 414 3747
E-mail: x.yao@cs.bham.ac.uk

For an application form, please contact:

Miss Helen Booth
Postgraduate Secretary for Research
Department of Civil Engineering
University of Birmingham
Tel: +44 (0)121 414 4160
Fax: +44 (0)121 414 5051
Email: h.r.booth@bham.ac.uk

The cover sheet for the application can be found here as a WORD file.

Project Outline:

Currently the type of maintenance applied to the railway track and its frequency is based on engineering judgement. This judgement is based on data recorded by measuring vehicles (known as geometry data) and the knowledge of the maintenance engineer responsible for a particular section of track. In practice several methods of maintenance may be used, including that related to the ballast (such as tamping and stone-blowing) and that concerning the subgrade or formation (such as improving the drainage or subgrade remediation methods). However, because of the complexities of the railway system, the inherent uncertainties associated with its performance, and the reliance of engineering knowledge of a particular section of track, which by its nature must be built up over many years, often inappropriate maintenance may be applied. To address this, the project proposes the development of a decision support system which would guide decisions regarding track maintenance. There are a number of systems, such as ECOTRACK, which may be used to assist with making maintenance and renewal decisions. However, recommendations for maintenance and renewal made by these systems are dependent solely on available data, they are used for lengthy sections of track (e.g. 250 m) and use maintenance standards and intervention levels to guide the decision making process. Consequently, their effectiveness is dependent on the availability and quality of data as well as the suitability of maintenance standards. Furthermore, by their nature these systems lend themselves to treating the symptoms rather than the cause of a particular problem with the consequent likelihood of a recurrence of the problem. The proposed system however, is intended for smaller problematic sections of track and will use not only recorded track data but will also incorporate the knowledge of track engineers. Thus the system will be able to direct the engineer to the cause of a particular problem rather than merely help to make a decision on the treatment of the symptoms.

The system would capture engineering knowledge, data recorded by track measuring vehicles and from track inspections to inform the engineer of the most likely sources of the problem(s) and suggest the most appropriate remediation technique(s).

The core of the system will use Computational Intelligence techniques including Evolutionary Computation and Neural Networks for studying the data recorded by track measurement vehicles. Additionally, data mining techniques will be employed to omit erroneous data from the decision making process. Evolutionary Computation is inspired by the process of evolution in nature and is a powerful technique to `evolve' models that can help solve complex problems. Evolutionary computation techniques work even when little or no information is available about the problem structure and no assumptions are made about the optimal model. Neural networks are powerful non-linear function approximators that can learn complex non-linear models from historical data. When combined, these techniques can help us not only to detect the cause of problems that occur on the track effectively but also understand the mechanism by which these factors interact and cause track failure. The successful candidate will be able to interact with members of CERCIA (http://www.cercia.ac.uk/) and the Natural Computation Group (http://www.cs.bham.ac.uk/research/NC/) at Birmingham.

NOTE: The studentship is only available to UK home / EU candidates.

DEADLINE: Friday 18th February 2005.