Premium
An initial evaluation of the detection and diagnosis of power plant faults using a deep knowledge representation of physical behaviour
Author(s) -
HERBERT M.R.,
WILLIAMS G.H.
Publication year - 1987
Publication title -
expert systems
Language(s) - English
Resource type - Journals
SCImago Journal Rank - 0.365
H-Index - 38
eISSN - 1468-0394
pISSN - 0266-4720
DOI - 10.1111/j.1468-0394.1987.tb00132.x
Subject(s) - computer science , fault detection and isolation , representation (politics) , fault (geology) , constraint (computer aided design) , operator (biology) , spurious relationship , prolog , artificial intelligence , data mining , machine learning , reliability engineering , mechanical engineering , biochemistry , chemistry , engineering , repressor , seismology , politics , political science , transcription factor , law , actuator , gene , geology
This paper describes an initial evaluation of a deep knowledge representation technique called Incremental Qualitative Analysis (IQA) to assess whether it is reasonably able to discriminate between different behaviours and hence different faults on a power plant. As an example the detection and diagnosis of faults in a PWR pressuriser sub‐system have been implemented in Prolog and evaluated using transients from a reference computer model of a pressuriser in conjunction with an input processor. The application differs from earlier systems for diagnosing faults in electronic hardware in that it is necessary to represent external feedback, two‐way components and sensor failures. The diagnosis is by synthesis and uses ‘constraint satisfaction and suspension’ techniques. For the present implementation all imposed faults have been successfully detected and diagnosed with reasonable determinacy, and a spurious detection has never occurred. IQA appears to be a suitable basis for knowledge‐based operator aids although to produce a real‐time aid appears to require more sophisticated techniques for fault detection and diagnosis, such as ‘constraint propagation’ and ‘candidate generation”. These techniques also offer the prospect of providing pseudo‐causal explanations to justify their hypotheses.