NEUROFUZZY APPROACHES AND THEIR APPLICATION TO NUCLEAR-POWER SYSTEMS

Neurofuzzy approaches (NFA) utilize a variety of neural and fuzzy syne rgisms that exploit a measured tolerance for imprecision and uncertain ty for the purpose of enhancing flexibility and tractability in models and systems. It is theoretically expected and empirically confirmed t hat neurofuzzy approaches when appropriately structured allow for impr oved control over the modeling economy or parsimony resulting in easie r to develop and modify systems. Hence, they hold considerable promise for significant enhancements in the control and safety of nuclear pla nt appurtenances, components and systems. Two nuclear power system app lications are presented in this paper. The first is in the reactor con trol area. It uses neural networks to predict power trajectories and f uzzy rules that incorporate such predictions in proactive or anticipat ory strategies in order to improve power maneuvers during reactor star tup. The second is in the area of safety,where neural mappings are use d to produce fuzzy values for epistemic variables. The methodology is extending the notion of measurement to variables with functional or op erational significance and hence is referred to as virtual measurement ; it is applied to flow visualization and holds considerable promise f br improving diagnostics and hence safety in nuclear reactors.