Stochastic auto-regressive reactor noise analysis for on-line vibration monitoring of core support barrel in PWR

Abstract

A stochastic autoregressive(AR) noise analysis program applicable to on-line vibration monitoring of PWR core is suggested. the main algorithm of this computer program is composed of two parts, the linear least square(LLS) calculation part and the Gauss-Newton iteration part, to obtain the vibration model parameters of reactor core. Specially, Gauss-Newton iteration part introduced in the second step of modeling process is based on the consideration of non-stationary conditions of operating reactor power. It was the first application in the field of PWR noise analysis. For the preparation of input data files, noise siganls are recorded from the ex-core neutron instrumentation channel in the ULJIN unit 1 unclear plant during full power operations. Then, with this analog signals, the analog-to digital(A/D) conversions are performed at a sampling interval of 100 Hz by 12 bit A/D conversion system. Those converted time series data are therefore used to obtain the AR power spectral densities (PSDs) in the frequency domain. The optimal model, which can represent on-line mechanical dynamics of ULJIN reactor core, is selected as AR(12) model by the Akaike Information Criterion(AIC) method and the F-test method; Both of the calculation results, the AIC result and the F-test method; Both of the same value in the estimation of optimal model order. therefore, the vibration information natural frequencies and damping ratios etc. of core support barrel(CSB) in the ractor vessel is derived from the AR(12) model equation subsequently. While, for the validation of this noise analysis program, conventional Fourier spectrum analysis is performed first. Through this analysis, it is verified that two patterns of noise spectra obtained by the AR method and the FFT method are nearly identical. Additionally, the mechanism of CSB vibration can also be estimated from the coherence analysis between excore detector singnals. Next, a comprehensive structrual analysis based on the mechanica...