A methodology for performing virtual measurements in a nuclear reactor system

A novel methodology is presented for monitoring nonphysically measurable variables in an experimental nuclear reactor. It is based on the employment of artificial neural networks to generate fuzzy values. Neural networks map spatiotemporal information (in the form of time series) to algebraically defined membership functions. The entire process can be thought of as a virtual measurement. Through such virtual measurements the values of nondirectly monitored parameters with operational significance, e.g., transient-type, valve-position, or performance, can be determined. Generating membership functions is a crucial step in the development and practical utilization of fuzzy reasoning, a computational approach that offers the advantage of describing the state of the system in a condensed, linguistic form, convenient for monitoring, diagnostics, and control algorithms