Thesis

Combining knowledge based systems and machine learning for turbine generator condition monitoring

Conteúdo disponível para baixar

Baixar PDF
Creator
Rights statement
Awarding institution
  • University of Strathclyde
Date of award
  • 2009
Thesis identifier
  • T13123
Qualification Level
Qualification Name
Department, School or Faculty
Abstract
  • This thesis reports on the design and development of a prototype condition monitoring system. The prototype system was developed for British Energy to assist the Rotating Plant and Dynamics Team in assessing the routine alarms triggered by their on-line condition monitoring system which continually monitor their turbine generators. The prototype comprises of two distinct modules. The first module is a Rule-Based Expert System which assesses the routine alarms using knowledge captured from the condition monitoring experts within the Rotating Plant and Dynamics Team. A Rule-Based Expert System approach was chosen so that there was a clear and transparent explanation provided with each assessment which allows the expert user to verify the result through following the assessment rationale. The second module is a learning assistant which was developed to assist the experts and knowledge elicitation engineer in capturing the explicit rule based knowledge used by a Rule-Based Expert Systems. This module uses a novel adapted version of the Machine Learning (ML) approach, Explanation Based Generalisation (EBG), to help derive knowledge from single training example and background causal behavioural knowledge of the turbine generator. This thesis outlines the rationale behind the selection of these approaches for the prototype system developed through a review of both Artificial Intelligence (AI) and ML approaches. A detailed description of the design approach and system architecture is given for both modules and a comprehensive review of the performance of both modules based on the results of system testing on genuine case study data is presented.
Resource Type
DOI
Date Created
  • 2009
Former identifier
  • 946602

Relações

Itens

Miniatura Título Data de carga Acesso Ações
file details PDF of thesis T13123 2021-07-02 Público Baixar