ABSTRACT
A novel approach is presented that couples Artificial Neural Networks (ANNs) with rule-based fuzzy expert systems for the purpose of monitoring nuclear reactor systems. A model-referenced approach is utilized which provides timely concise and task specific information about the state of the system. In this approach, the rule-based system performs the basic interpretation and processing of the model data, and pre-trained ANN’s provide the model. Having access to a set of neural networks that typify general categories of system behavior, the rule-based system performs identification functions. The output of the neural networks is membership functions which condense information about the state of a system in a form convenient for a rule-based system. This allows the identification function to be performed in a time frame faster or at least comparable to the temporal evolution of the system. The methodology is demonstrated using actual data from an experimental nuclear reactor exhibiting excellent robustness to noisy as well as faulty signals.
