Neural Networks on Parallel Systems-Logic Programming Implementation and Medical Image Analysis

NNPS&LPIMIA - ITDC-106

Keywords: neural networks, parallel processing, logic programming, robotics, fault diagnostics

Start Date: 1 February 95 / Duration: 30 months

Objectivies and Approach

The objective of this project is to develop, analyse and implement computation-intensive algorithms for real time control and fault diagnostic applications on a scaleable computer. Specifically, the real time control of robotic manipulators and process condition monitoring are targeted. The workplan includes analysis and synthesis of parallel implementations of neural network computations, analysis of robot trajectory control and implementation of scaleable robotic algorithms and an exploratory study on a parallel implementation of a multisensor tool condition monitoring system.

Progress and Results

Two neural network paradigms were investigated: a multilayer perceptron for the neural controller implementation, and adaptive resonance theory for the fault diagnostic implementation. Multilayer perceptrons are well-known for control applications, whereas adaptive resonance theory is ideal for feature extraction jobs in fault monitoring and diagnosis

Work is going on regarding the experiment for a real time robot control system, as well as on the fault diagnostic application.

Coordinator

University of Malaya
Department of Physics
Faculty of Computer Science and
Information Technology
50603 Kuala Lumpur, MY

EU Partners

Telmat Multinode, F

Non-EU Partners

University of Malaya, MY
Kuala Lumpur Hospital, MY

CONTACT POINT

W.A.T. Wan Abdullah
Tel: +60 3 755 54 66/ 60 3 759 41 92/ 60 3 759 42 06
Fax: +60 3 757 36 61/ 60 3 759 41 46
E-mail: wat@cc.um.edu.my