Use this URL to cite or link to this record in EThOS:
Title: Intelligent monitoring of a complex, non-linear system using artificial neural networks
Author: Weller, Peter Richard
ISNI:       0000 0001 2430 8091
Awarding Body: City University London
Current Institution: City, University of London
Date of Award: 1997
Availability of Full Text:
Access through EThOS:
Access through Institution:
This project uses advanced modelling techniques to produce a design for a computer based advisory system for the operator of a critical, complex, non-linear system, typified by a nuclear reactor. When such systems are in fault the operator has to promptly assess the problem and commence remedial action. Additional accurate and rapid information to assist in this task would clearly be of benefit. The proposed advisory system consists of two main elements. The plant state is determined and then the future condition predicted. These two components are linked by a common data flow. The diagnosed condition is also used as input for the predictive section. Artificial Neural Networks (ANNs) are used to perform both diagnosis and predictions. An ANN, a simplified model of the brain, can be trained to classify a set of known inputs. It can then classify unknown inputs The predictive element is first investigated. The number of conditions that can be predicted by a single ANN is identified as a key factor. Two distinct solutions are considered. The first uses the important features of the fault to determine an empirical relationship for combining transients. The second uses ANNs to model a range of system transients. A simple model is developed and refined to represent an important section of a nuclear reactor. The results show good predicted values for a extensive range of fault scenarios. The second approachis selected for implementation in the advisory system. The diagnostic element is explored using a set of key transients. A series of ANNs for diagnosing these conditions are developed using a range of strategies. The optimum combination was selected for implementation in the advisory system. The key plant variables which contributed most to the ANN inputs were identified. An implementation of the advisory system is described. The system should be a single suite of programs with the predictive and diagnostic sections supported by a controller module for organising information. The project concludes that the construction of such a system is possible with the latest technologies.
Supervisor: Not available Sponsor: Not available
Qualification Name: Thesis (Ph.D.) Qualification Level: Doctoral
EThOS ID:  DOI: Not available
Keywords: T Technology Automatic control Control theory Bionics