Knowledge based fault monitoring for large complex systems
Knowledge based fault monitoring for large complex systems
The application of the technology of Knowledge-Based Systems is being considered for real time fault monitoring of large complex dynamic systems. Two of the most critical problems in such monitoring systems are the perception or detection of system abnormality and diagnosis of system faults responsible for the abnormality. Although conventional analytical techniques such as system redundancy approaches have proved useful in solving these problems, the most positive design of the monitoring functions which includes system failure detection, fault isolation and identification, and system recovery decisions and actions usually requires the merging of analytical evaluation and logic reasoning, where the combination of a variety of sources of knowledge from human expertise to system structural and functional information and analytical models of systems is indispensible and most effective. This has led to an Intelligent Qualitative and Quantitative Reasoning Scheme as a problem solution architecture. This thesis describes research to investigate knowledge based techniques to the problems of failure detection, fault source isolation and fault pattern identification of large complex systems. It first provides a discussion of the problem domain of system fault monitoring in a general manner; then it presents the specific methodologies for above critical problems using the XB-70 aircraft flight control system as a test model and the KEE systems as an implementation tool. An important part of the work has been to develop an effective rule based constraint evaluation approach to fault isolation problems. The approach is essentially algorithmic and makes use of hypothesis generate and test paradigm and the concept of `constraints'. However, certain heuristic rules can be easily incorporated such that the entire methodology can deal with a wide range of failure cases in an efficient way. Finally some of the remaining problems related to the current research and the subjects as extensions to the present research are forseen.
University of Southampton
1991
Xu, Yuan Ming
(1991)
Knowledge based fault monitoring for large complex systems.
University of Southampton, Doctoral Thesis.
Record type:
Thesis
(Doctoral)
Abstract
The application of the technology of Knowledge-Based Systems is being considered for real time fault monitoring of large complex dynamic systems. Two of the most critical problems in such monitoring systems are the perception or detection of system abnormality and diagnosis of system faults responsible for the abnormality. Although conventional analytical techniques such as system redundancy approaches have proved useful in solving these problems, the most positive design of the monitoring functions which includes system failure detection, fault isolation and identification, and system recovery decisions and actions usually requires the merging of analytical evaluation and logic reasoning, where the combination of a variety of sources of knowledge from human expertise to system structural and functional information and analytical models of systems is indispensible and most effective. This has led to an Intelligent Qualitative and Quantitative Reasoning Scheme as a problem solution architecture. This thesis describes research to investigate knowledge based techniques to the problems of failure detection, fault source isolation and fault pattern identification of large complex systems. It first provides a discussion of the problem domain of system fault monitoring in a general manner; then it presents the specific methodologies for above critical problems using the XB-70 aircraft flight control system as a test model and the KEE systems as an implementation tool. An important part of the work has been to develop an effective rule based constraint evaluation approach to fault isolation problems. The approach is essentially algorithmic and makes use of hypothesis generate and test paradigm and the concept of `constraints'. However, certain heuristic rules can be easily incorporated such that the entire methodology can deal with a wide range of failure cases in an efficient way. Finally some of the remaining problems related to the current research and the subjects as extensions to the present research are forseen.
This record has no associated files available for download.
More information
Published date: 1991
Identifiers
Local EPrints ID: 461008
URI: http://eprints.soton.ac.uk/id/eprint/461008
PURE UUID: 534244a2-61c0-4414-8ddd-dfb38d0b7ec1
Catalogue record
Date deposited: 04 Jul 2022 18:33
Last modified: 04 Jul 2022 18:33
Export record
Contributors
Author:
Yuan Ming Xu
Download statistics
Downloads from ePrints over the past year. Other digital versions may also be available to download e.g. from the publisher's website.
View more statistics