Research On The Mass Estimation Of Loose Parts In Nuclear Power Plant

Based on the National Natural Science Foundation of China "Study on weak burst signal detection and extraction(No.51175466)" and the High Technology Research and Development Program of China "Research on the Key Technology of Loose Part Monitoring in Nuclear Power Plant (No.2009AA04Z426)", the thesis titled as "Research on the Mass Estimation of Loose Parts in Nuclear Power Plant" is proposed, researches of detecting and discriminating the loose part impact signal, estimating the mass of the loose part are carried out.In chapter1:The neccessasity of the role that Loose Part Monitoring System (LPMS) plays in the Nuclear Power Plant (NPP) is illustrated. The research progress and the existed products of LPMS are introduced. Research frame of this thesis is proposed based on the summarized methods of mass estimation of loose part.In chapter2:To the difficulty of detecting weak impact signal in strong background noise, a method for weak impact signal detection and discrimination based on Para-approximation entropy (P-ApEn) is proposed. Pre-process method using Auto Regressive (AR) model and Teager Energy Operator (TEO) is used to enlarge the detectable range of noisy signal. Para-Approximation Entropy (P-ApEn) is proposed to establish the discrimination model for weak impact signal, which can effectively discriminate the impact signal and other interferences.In chapter3:Based on the feature of Marginal Hilbert Spectrum (MHS), a method for mass estimation is proposed. Analysis of simulation signals show that MHS is capable to describe the frequency content more precisely than Fourier spectrum for loose part impact signal. According to the feature of MHS and the characteristics of mass, Support Vector Classification (SVC) and Regression (SVR) are combined to establish the compound sub-regression model for mass estimation. The experimental data analysis of the proposed method proves that the compound method can estimate mass value more precisely.In chapter4:A mass estimation method based on improved time duration estimation and Hertz’s contact theory is proposed based on time-frequency distribution of impact signal. According to the corresponding relation between the amplitude and its frequency, estimation of the contact duration time is improved for reducing the difficulty of correctly estimating the contact time in Hertz’s theory, and the attenuation of impact signal amplitude is amended by utilizing Hankel function..In chapter5:According to the fact that the frequency content of impact signal can be affected by impact velocity, a method based on S transform is proposed to estimate the mass of loose part. The affect patterns led by impact velocity and impact location are analyzed qualitatively. Compared with the other regular time-frequency method in processing the periodic and non-stationary simulation signal, S transform is proven to be rational for impact signal time-frequency description. The dimension of time-frequency spectrum is reduced by segmental averaging and the mass feature is extracted using frequency slicing of the time-frequency spectrum.In chapter6:Mass estimation prototype system is developed using Matlab and SQL Server based on the analysis of current LPMS products. System design and function assignment is proposed, the main functional module and the user interface are introduced. Data from simulation experiment are processed and analysed using the system.In chapter7:The research content of the thesis are summarized and the future of LPMS and mass estimation development is prospected.