Uncertainty modeling in anticipatory systems [nuclear reactor operator emulation]

An anticipatory system is a system containing an internal predictive model of itself and/or its environment which allows it to change state at a time t according to the model´s prediction pertaining to a time t+Δt later. The anticipatory paradigm offers a useful framework for the development and integration of automated monitoring, diagnostic and control functions that may enhance the safety margins and overall performance of large complex systems. In the research described, a rule-based system emulates aspects of the monitoring, diagnostic and control behavior of a nuclear reactor operator. It receives data from sensors and uses a predictive model that simulates the reactor´s behavior. Fuzzy information granules provide a syntax for representing facts and inferring current as well as anticipated performance of components, and subsystems. The basic assumption of the anticipatory paradigm as employed in this research is that a diagnostic and/or control decision can be based on a performance measure; and that it is desirable to estimate future performance and incorporate it in the control strategy. The methodology presented allows for estimating present as well as future performance on the basis of probabilistic and possibilistic information and is demonstrated with the simulation of an anticipatory system