Nuclear Science and Engineering · 1979 · 16 citations · 0 references
EngineeringNuclear EngineeringSystems EngineeringModel-based Systems EngineeringProcess DynamicsSystem DynamicsModel IdentifiabilityProcess IdentificationReactor DynamicsReactor ModelingState EstimationSystem IdentificationDynamic Modeling Of Nuclear ReactorsEmpirical EvaluationNoise AnalysisNoise Analysis In ReactorsNumerical IdentifiabilityParameter Estimation In Reactor ModelsParameter IdentifiabilityReactor Thermal-hydraulic Modeling
Reactor noise analysis using autoregressive modeling is investigated for system identifiability, especially useful when prior knowledge of process dynamics is limited. The study proposes a practical RRV checking method to detect when only part of an identified AR model is meaningful. The authors derive a condition identifying when only part of an AR model is meaningful and apply the RRV checking method to models built from experimental data of the Japan Power Demonstration Reactor II. The RRV checking method confirmed that models previously deemed unacceptable due to physically implausible components were actually acceptable, identifying fuel‑temperature–neutron‑density and steam‑control‑valve–reactor‑pressure responses as reliably observable from reactor noise.
The feasibility of reactor noise analysis by autoregressive (AR) modeling is studied from the viewpoint of system identifiability. A condition is derived in which only a part of the identified model becomes meaningful. A practical checking method termed “RRV checking” is proposed, with which the occurrence of the condition is recognized a posteriori for the estimated AR model. This method is applied to AR models obtained by processing the experimental data from the Japan Power Demonstration Reactor II. These models would have been discarded from a conventional viewpoint, since some parts of the model showed physically implausible characteristics. It is verified that the RR V checking method and the empirical evaluation of the usability of the model resulted in the same conclusion about the acceptability of the parts of the models. The processes evaluated to be identifiable from the reactor noise are the response of the fuel temperature to the neutron density and the response of the steam control valve to the reactor pressure. The present method is particularly useful if a priori knowledge about the dynamics of the objective process is limited before the identification experiment.