Reactor control and transient identification by neural networks using wavelets and time-frequency atoms.
Committee ChairWilliams, J. G.
MetadataShow full item record
PublisherThe University of Arizona.
RightsCopyright © is held by the author. Digital access to this material is made possible by the University Libraries, University of Arizona. Further transmission, reproduction or presentation (such as public display or performance) of protected items is prohibited except with permission of the author.
AbstractIn this dissertation, techniques that have potential applications to the digital operation of nuclear reactors are developed. A digital controller is developed to plan feasible control actions and digital transient identifiers are developed to detect unexpected events. Given an operational objective, the model based controller identifies a sequence of control actions that will fulfill the operational objective. The hazard anticipator checks the feasibility of the control. Only by means of the hazard anticipator, can the digital control system safely perform control actions without violating technical specifications and the limits of the physical system. In the controller, a precursor population meter implemented as a finite filter of the power history provides effective and accurate estimation of the delayed neutron precursor population. In this dissertation, techniques of transient identification are developed to detect unexpected events. Neural networks are used to identify different types of transients based on the distinct features extracted from the transients by matching pursuit decomposition or by shiftable wavelet transformation. The techniques have been shown to work well in extensive tests.
Degree ProgramNuclear and Energy Engineering