On-line stability monitoring of BWR's using artificial neural networks

The problem of uncovering stability parameters of Boiling Water Reactors (B WRs) from short neutron noise records for on-line monitoring is investigate d. The aim is to use the smallest possible number of points of the auto-cor relation function (ACF) at the shortest possible time lags as input to a nu mber of backpropagation artificial neural networks (BP ANNs), which are tra ined to approximate the relationship that exists between the ACF of the neu tron noise signals and the stability parameters of interest (e.g. the decay ratio (DR)). Tests on novel ACFs illustrate the potential of the proposed approach for on-line stability monitoring. (C) 1999 Elsevier Science Ltd. A ll rights reserved.