Skip to main content
U.S. Department of Energy
Office of Scientific and Technical Information

Virtual measurements using neural networks and fuzzy logic

Conference · · Proceedings of the American Power Conference; (United States)
OSTI ID:7116959
; ;  [1]
  1. Tennessee Univ., Knoxville, TN (United States). Dept. of Nuclear Engineering

This paper reports on a fuzzy-neural methodology which is developed for the purpose of measuring plant parameters (e.g., performance) difficult to estimate in a timely and reliable manner. In the virtual measurement methodology presented the output of a measurement is considered to be a fuzzy number, called Virtual Measurement Value (VMV). VMVs are obtained through a complex-to-simple mapping performed by artificial neural networks. Each virtual measurement value is uniquely identified by a trapezoidal membership function learned by a simple network using the back-propagation learning algorithm. The actual output at any given time is selected through an index of dissemblance, which is a measure of similarity between two fuzzy numbers. The methodology is tested with start-up data form an experimental nuclear reactor.

OSTI ID:
7116959
Report Number(s):
CONF-920432--
Journal Information:
Proceedings of the American Power Conference; (United States), Journal Name: Proceedings of the American Power Conference; (United States) Vol. 54:2; ISSN PAPWA; ISSN 0097-2126
Country of Publication:
United States
Language:
English