Expert systems in A.C. induction motor fault diagnosis.

Recent research has produced reliable techniques for non-invasive, on-line monitoring and fault diagnosis of large induction motors. Human knowledge and experience has to be used in conjunction with these techniques in order to circumvent the problem of unavailable or uncertain data and to make judgement on the degree of severity of any discovered faults. The use of knowledge based system programming techniques allows this process to be automated so reducing the cognitive load of human expert diagnosticians and making this knowledge more widely available where motor failures can result in serious economic or safety implications. From a detailed analysis of the sources and nature of such knowledge, a novel knowledge base architecture is proposed which provides effective and efficient representation. The key interrelated components of this knowledge base are a fault and symptom orientated hierarchical tree of frames and data base of case histories of motor faults. The architecture, implemented in Prolog and C programming languages, provides a graphical user interface and direct connection to data gathering/analysis equipment for real-time (relative to the time constraints of the diagnostic process), on-line operation. The performance of this configuration has been evaluated as adequate for the effective prognostication of induction motor electromechanical faults. A scheme for inductive learning of new information from case histories is also reported. Based on the ID3 learning algorithm, the approach determines emerging patterns and relationships as more diagnoses are made. A method of quantifying the generality of the induced conclusions is also defined and described.

Author dc:creator

• Leith, Douglas

Advisor dc:contributor.advisor

• N.D. Deans and W.T. Thomson

• Induction motor fault diagnosis
• Knowledge-based systems
• Real-time condition monitoring
• Fault and symptom hierarchies
• Case-based reasoning
• Inductive learning

Language dc:language

en

Identifier

oai:rgu-repository.worktribe.com:3307184
https://doi.org/10.48526/rgu-wt-3307184

OAI identifier oai:identifier

oai:rgu-repository.worktribe.com:3307184