Methodology for developing second-generation expert systems using qualitative simulation
There are certain shortcomings with the current generation of expert systems that must be overcome. The first limitation is that they tend to break down when faced with unforeseen circumstances at the periphery of their knowledge. The second problem is that they cannot explain the underlying reasoning. Basically, these shortcomings are due to the shallow reasoning employed in the construction of many expert systems. The typical knowledge base does not support reasoning from first principles. This research presents an alternative approach to building the knowledge base using qualitative simulation of the physical system of interest. A six-step methodology is described that constructs a diagnostic ES from specifications of the physical system. A key concept in the methodology is the creation of a qualitative model, using a network of expert systems. In this network, the behavior of each component of the physical system is defined within an ES node. The results of running the qualitative model are used in conjunction with reliability and cost data to produce the diagnostic ES. The resulting ES, therefore, employs a deep reasoning based on the underlying physical mechanism rather than rules-of-thumb from a human expert.
- Research Organization:
- Arizona State Univ., Tempe (USA)
- OSTI ID:
- 6210375
- Country of Publication:
- United States
- Language:
- English
Similar Records
Methodological approach to a re-usable fuzzy expert system
Development of a methodology for knowledge elicitation for building expert systems