Anticipatory systems using a probabilistic-possibilistic formalism

A methodology for the realization of the Anticipatory Paradigm in the diagnosis and control of complex systems, such as power plants, is developed. The objective is to synthesize engineering systems as analogs of certain biological systems which are capable of modifying their present states on the basis of anticipated future states. These future states are construed to be the output of predictive, numerical, stochastic or symbolic models.The mathematical basis of the implementation is developed on the basis of a formulation coupling probabilistic(random) and possibilistic(fuzzy) data in the form of an Information Granule. Random data is generated from observations and sensors input from the environment. Fuzzy data consists of epistemic information, such as criteria or constraints qualifying the environmental inputs. The approach generates mathematical performance measures upon which diagnostic inferences and control functions are based. Anticipated performance is generated using a fuzzified Bayes formula. Triplex arithmetic is used in the numerical estimation of the performance measures. Representation of the system is based upon a goal-tree within the rule-based paradigm from the field of Applied Artificial Intelligence. The ensuing construction incorporates a coupling of Symbolic and Procedural programming methods.As a demonstration of the possibility of constructing such systems, a model-based system of a nuclear reactor is constructed. A numerical model of the reactor as a damped simple harmonic oscillator is used. The neutronic behavior is described by a point kinetics model with temperature feedback. The resulting system is programmed in OPS5 for the symbolic component and in FORTRAN for the procedural part. Examples are used to demonstrate the use of the proposed approach for the construction of engineering analogs of anticipatory biological systems. It is shown how anticipatory systems can use measures of performance to represent the current as well as anticipated state in such a manner that decisions about changing state are related to a search for maximizing the performance associated with a state variable.U of I OnlyETDs are only available to UIUC Users without author permissio