Qualitative process modeling for expert system fault diagnosis

Da-zhuang, Guo (1992). Qualitative process modeling for expert system fault diagnosis PhD Thesis, School of Engineering, University of Queensland. doi:10.14264/uql.2016.437

Attached Files (Some files may be inaccessible until you login with your UQ eSpace credentials)
Name Description MIMEType Size Downloads
THE7640.pdf THE7640.pdf application/pdf 2.58MB 0
Author Da-zhuang, Guo
Thesis Title Qualitative process modeling for expert system fault diagnosis
School, Centre or Institute School of Engineering
Institution University of Queensland
DOI 10.14264/uql.2016.437
Publication date 1992-01-01
Thesis type PhD Thesis
Supervisor Robert B. Newell
Total pages 197
Language eng
Subjects 09 Engineering
Formatted abstract
As shallow knowledge expert systems have entered their application stage, their limitations have begun to be revealed. In order to compensate for these drawbacks more attention has been recently focused on combining deep knowledge with shallow knowledge in expert systems. 

This thesis investigates the role of qualitative process modeling in expert system fault diagnosis and proposes a deep knowledge qualitative approach which incorporates two deep knowledge techniques, the signed directed graph and the qualitative simulation with QSIM, to an expert system for process fault diagnosis based on the hypothesis-and-test strategy. 

Improvements are made in both qualitative techniques in this thesis. In the signed directed graph, there are two kinds of backward reasoning algorithms based on the ideas of the maximal strongly connected component and on the consistent rooted tree respectively. This thesis gives an example to show the differences between the results from these two algorithms although it was said that the two algorithms produced the same result. The improved algorithm is developed based upon both the maximal strongly connected component and the consistent rooted tree. The qualitative simulation technique is still at an early stage of development and can be applied only to simple systems. Two new constraints and one new global filter are defined and applied to remove some ambiguous and spurious behaviours from the gualitative simulation. 

The improved algorithm mentioned above is coded into the PC-based program DIGRAPH in PASCAL for search in the signed directed graph, and the PC-based program CQSIM is written in C for the gualitative simulation with QSIM. The interface subroutines are developed to link both programs with a PC-based expert system development environment, Intelligence Compiler, and a prototype of the integrated expert system for fault diagnosis is developed to combine the deep knowledge gualitative system based on the proposed approach with the shallow knowledge base. 

Models of a cascade water tank system and a pipeline system are developed both in the signed directed graph and in qualitative simulation with QSIM and used to demonstrate fault diagnosis with the proposed deep knowledge expert system. The models of the pipeline system are used to show that the expert system can still work even when one of the two techniques cannot deal with some system failures.
Keyword Expert systems (Computer Science)
Fault location (Engineering)

Document type: Thesis
Collection: UQ Theses (RHD) - UQ staff and students only
Citation counts: Google Scholar Search Google Scholar
Created: Wed, 08 Jun 2016, 21:34:26 EST by Ms Dulcie Stewart on behalf of Information Systems and Resource Services, Library