Hybrid expert system; Neural network methodology for transient identification

This paper presents a methodology that couples rule-based expert systems using fuzzy logic, to pre-trained artificial neutral networks (ANN) for the purpose of transient identification in Nuclear Power Plants (NPP). In order to provide timely concise, and task-specific information about the may aspects of the transient and to determine the state of the system based on the interpretation of potentially noisy data a model-referenced approach is utilized. In it, the expert system performs the basic interpretation and processing of the model data, and pre-trained ANNs provide the model. having access to a set of neural networks that typify general categories of transients, the rule based system is able to perform identification functions. Membership functions - condensing information about a transient in a form convenient for a rule-based identification system characterizing a transient - are the output of neural computations. This allows the identification function to be performed with a speed comparable to or faster than that of the temporal evolution of the system. Simulator data form major secondary system pipe rupture is used to demonstrate the methodology. The results indicate excellent noise-tolerance for ANN's and suggest a new method for transient identification within the framework of Fuzzy Logic.