|
Project Proposals for William Flynn Scholarship:
IIndex
page
Project Number: 3
Project Title: Hybrid Intelligent Techniques for the rapid test
and diagnosis of electronic systems
Project Supervisor: Dr. Liam Maguire, Professor Martin Mc Ginnity
Background
As electronic systems increase in complexity, the need for automated
diagnostic tools has become more acute. This is exacerbated by reduced
time-to-market, and shorter product lifecycles, leading to little
development time being available for diagnostics. Although much
research has been carried out in the area, much remains to be done,
particularly in the deployment of useful tools in real applications.
For example, most complex electronic systems are now microprocessor
or DSP driven which involve the tight integration of hardware and
software, and therefore present additional problems.
Initial research by the ISEL group have identified that if intelligent
techniques are to be widely used for diagnosis, then a number of
issues must be addressed.
1. Development tools must make the intelligent techniques transparent
to the user; for example all representations must be entered in
commonly understood formats.
2. Knowledge must be reusable, in order to make development times
short. For example, knowledge learned in diagnosing one product
must be generalisable so it can be used on newer products, or tests
already available for pass/fail testing can be reused for diagnosis.
3. Approaches must overcome the knowledge acquisition bottleneck.
This research has proposed a hybrid solution that attempts to mimic
the "average" technician and restrict the development
of new diagnostic solutions to a minimum. This results in a tradeoff
between development time and a diagnostic accuracy that can be acceptable.
The approach has developed a hybrid strategy incorporating both
model and case based reasoning techniques. Initially a dependency
model is required, in the form of a block diagram which highlights
the main components within the system and the inherent interconnections.
The creation of this model demands no aprior knowledge of intelligent
techniques and the application of test and diagnostic scenarios
automatically results in the creation of a case history for the
device under test. This case history is then used to represent the
experience of a test technician as it determines the most likely
cause of a fault based on the probabilities of previous occurrences.
Similarly this knowledge can be used as an initial weighting for
a new case history when the system moves to consider an alternative
product. As a result the approach provides a very transparent system
to the user as knowledge elicitation occurs automatically with its
operation.
There are a number of methods to extend and improve this intelligent
diagnostic strategy so that it can mimic a more expert technician.
For example, the initial approach to reasoning across the case history
uses probablistic reasoning based on the statistical occurrence
of the diagnosis. This could be readily extended to fuzzy reasoning
to mimic a more typical human technician and this would also integrate
the inherent uncertainty associated with the process. In addition,
the application of neural networks would enable the strategy to
integrate an element of learning the relationships between faults
and diagnosis to further improve the final result. Finally evolutionary
algorithms could be applied to tune both the fuzzy and neural techniques
deployed and indeed the basic rules generated from the initial dependency
model.
Aims and objectives
This project attempts to further develop a hybrid intelligent strategy
for the test and diagnosis of electronic systems. The main objectives
of the project can be summarised as:
1. Provide a critical review of the various intelligent and hybrid
intelligent techniques for test and diagnosis of electronic systems.
2. Develop new approaches to how these techniques can be integrated
to provide a fast deployment system that is transparent to the user
and which mimics the expert technician.
3. Design and implement these approaches.
4. Verify and validate the research by implementing a number of
real cases from industry.
5. Improve the design and architecture by review and iteration.
Intellectual Challenge and training and benefits to the student
There are two main intellectual challenges relevant to the project.
The first involves the assessment of the various intelligent and
hybrid intelligent techniques that could be applied in this application
domain. The second involves the actual implementation of the selected
strategy and an assessment of its effectiveness, which necessitates
a high competence in programming and system design and development.
There is obviously a strong training component in the project in
terms of the differing intelligent techniques, assessment and analytical
techniques, and the application domain. The student will benefit
from the project in terms of his/her academic contribution to the
emerging disciplines of intelligent systems and test and diagnosis
but also by its application to real world problems.
Anticipated Outcomes
The successful completion of the project will demonstrate the effectiveness
of a hybrid intelligent architecture for the rapid test and diagnosis
of electronic systems. The academic contribution will be disseminated
in a number of academic journals such as IEEE Intelligent Systems,
Design and Test, Fuzzy Systems, Neural Networks, Systems Man and
Cybernetics.
If you are interested in being considered for a studentship please
contact
the Group Director, Professor T.M. McGinnity by email:
tm.mcginnity@ulst.ac.uk
or telephone: +44-(0)28-71375417.
See the current research section of this website
for details on research projects pursued by existing PhD students
| 