|
The outcomes of this project will be integrated with that of the
'health card' project, with prototype 'health card' pictured.
|
The monitoring of rail track conditions and identification of
‘bad track’ sections is traditionally undertaken using
a special Track Recording Vehicle (TRV), equipped with a variety of
sensors. However, while TRVs provide valuable information, they do
not necessarily provide all the information necessary for an
optimal monitoring and maintenance approach.
An alternative approach, taken up under this project, is to develop
fast and easy-to-use mathematical models that predict wagon
behaviour based on track geometry data. This project, which
commenced in late 2005, aims where possible, to combine outputs and
data from Projects 17 and 26 under the Theme 1 research programme,
and develop a method based on neural networks. This will be used to
identify track faults, programme maintenance and set speed
restrictions. A key outcome from the project will be neural network
models that will predict wagon performance-data based on given
track geometry car measurements. These models will allow a software
package to be developed that will calculate the wagon-track
performance over long lengths of track (i.e. whole haulage routes)
without the need for traditional and time-consuming wagon
simulation.
As a second focus of research, the project also aims to investigate
the possibilities offered by applying intelligent system
classification to data obtained from the Rail CRC "Health Card"
(Project 2), and systems that produce performance data for the
identification of track signatures of known track and wheel
defects.
Specific benefits of the project’s outcomes are expected to
include reduced maintenance, improved safety due to improved train
driving strategies (i.e. safe speeds) over critical sections of
track and reduced need to use Track Recording Vehicles. The project
was evaluated by STEM Partnerships in 2006 to have the potential to
deliver an expected value of $14.6 million over the next 15 years,
taking into account risks involved in delivering the technology.
Progress to date:
- Literature review undertaken.
- Preliminary programming work undertaken.
Future outcomes:
- Development of neural network model for separation of the normal
data from the exception data.
- Development of neural network model for classification of
exceptions with similar characteristics into useful groups for
reporting and maintenance planning.
Project Leader: Dr Colin Cole (Central Queensland University)
Project Manager: Mr Clive Plunkett (Queensland Rail)
