Improved PWR-neutron noise interpretation based on detailed vibration analysis

General Regression Neural Networks for the Concurrent, Timely and Reliable Identification of Detector Malfunctions and/or Nuclear Reactor Deviations from Steady-State Operation

“…NN Signals and NDs -S3K Signal Generation NN signals constitute the prevalent source of information for performing NR system analysis and identification, as well as real-time characterization of the NR operation mode and status. Further to the pure neutron noise part -which is inherent in the NN signal per se and considered crucial for the purposes of system verification as well as for on-line monitoring -, these signals may also encompass signal drifts and/or more systematic oscillations, which result from such (common, yet crucial) factors [2] as (a) core barrel beam mode, (b) cylindrical component shell mode, (c) cylindrical component mode, (d) fuel assembly beam mode 2 , as well as other temporary and/or harder to define sources of signal distortion, thus effectuating -at times -a significant corruption of the original NN signal to be analyzed. It is important that the various source(s) of NR signal corruption be distinguished and characterized as:…”

“…properties which are necessary for concurrently providing not only timely and reliable decisions on signal validity, but -furthermore -on the identity/cause of the driving perturbation. This contribution is organized as follows: Section II introduces the general framework (problem definition/ representation and employed signals/methodologies) of NR monitoring and NF anomaly/ND malfunction detection [2]; Section III implements signal verification and prediction via both parametric polynomial approximation (PA) [3], semiparametric splines (SPSs) [4] and non-parametric general regression artificial neural networks (GRNNs) [5] approaches on data that has been created using established simulation codes (e.g. [6][7]) and handled in its entirety as well as piece-wise (systematically derived from 10-fold cross-validation (CV) [8]), thereby establishing the validity and generalization properties of the three approximations methodologies on the present problem; Section IV critically presents and compares the verification/prediction results, with Section V summarizing the findings and putting forward potential extensions of the present research aims concerning NR monitoring and diagnosis.…”

Advances in automated noise data acquisition and noise source modeling for power reactors

“…In anticipation of a need relative to the Sequoyah noise data acquisition project",' one^anc two«-dimensional space-and energy-dependent neutron kintetics calculations were performed in the frequency domain to determine the response functions of ex-core neutron detectors to various postulated noise sources within the reactor core. In particular, since vibrations of PWR fuel assemblies have been observed (Wach and Sunder, 1977;Mayo and Currie, 1977) through their influence on the noise signals received by ex-core neutron detectors, there was interest in quantifying the amplitude of these vibrations. To do so, the spatial sensitivity function of the detectors, as well as the neutron flux perturbation resulting from the mechanical vibrations, must be known.…”