ON CAUSALITY ANALYSIS OF NUCLEAR REACTOR NOISE USING PARTIAL DIRECTED COHERENCE

摘要

This paper represents a first application of a causality technique based on the so-called partial directed coherence (PDC), widely applied in neuroscience and econometrics, to nuclear reactor signal noise analysis with the aim to examine and investigate the information flow paths and the cause-and-effect connections within a multivariate system, i.e. among different reactor system signals, e.g. ex/in-core neutron detectors and process signals. In this context, a PDC-based causality analysis algorithm has been implemented and several verification analyses have been carried out in order to qualify such technique for causality problems in nuclear power plants (NPP). To this aim, the algorithm has been verified first for a simple analytical model then it was applied to a simulation of a boiling water reactor (BWR) problem using the transient nodal code SIMULATE-3K (S3K), where a perturbation in the feedwater flow variable has been introduced. The goal is to illustrate the capabilities of this technique to capture the source of the system fluctuations and also to investigate how the perturbation is propagating within the system. In addition, numerical noise disturbances observed in a TRACE-S3K steady-state calculation were analyzed with the PDC-based causality method and results show a big potential of such technique in indicating the sources of possible numerical instabilities of complex systems. Finally, a real stability event, occurred in a Swiss NPP, has been studied with the current technique in order to identify the cause of the observed oscillations at a certain nominal frequency, and results show the potential of the technique to indicate possible rout-of-cause relationships between the examined signals. The current qualification of this technique in nuclear reactor noise analysis provides thereby confidence for its capabilities in signal diagnostics and causality studies.