Título: A methodology for generation of fault diagnostic knowledge
Autores: Sandrasegaran, Kumbesan
Fecha: 1994
Publicador: McGill University - MCGILL
Fuente:
Tipo: Electronic Thesis or Dissertation
Tema: Engineering, Electronics and Electrical.
Descripción: This dissertation presents a methodology for generation of fault diagnostic knowledge from a description of a device. The generated knowledge is to be used in a computer based learning environment for fault diagnostic tasks. In the past, developers of such environments obtained diagnostic knowledge through a knowledge engineering exercise with human experts. There are a number of drawbacks associated with such an approach. A major bottleneck in the development of such learning environments was capturing human expertise for diagnosis of an application, encoding this knowledge in a suitable form in a computer, and then verifying this knowledge. This exercise also depends on the availability and cooperation of a knowledgeable human expert(s). If more than one expert participates in the knowledge acquisition process, one may obtain contradictory information. Furthermore, most of the diagnostic knowledge is highly application specific thus making it useless for other applications. The starting point of the methodology of this thesis is a description of a device in terms of the components, component behaviors, and interconnections between components (structural knowledge). The end point of the methodology is a set of rules that can be used to diagnose faults in the device. The intermediate points are a behavioral and causal descriptions of the device, and a set of domain independent diagnostic strategies. This methodology has been applied to a counter circuit in the domain of digital electronics to test both the ability of the fault diagnostic system to diagnose faults in a device as well as to test its efficiency. The rules generated were able to successfully detect all the faults that were inserted in the counter application. Furthermore, as more diagnostic strategies were included in the diagnostic rule generation, the efficiency of the diagnostic system improved considerably.
Idioma: en