期刊
PROGRESS IN NUCLEAR ENERGY
卷 79, 期 -, 页码 8-21出版社
PERGAMON-ELSEVIER SCIENCE LTD
DOI: 10.1016/j.pnucene.2014.10.013
关键词
Sensor fault detection; Nuclear plant operations and maintenance; Bayesian artificial neural network (BANN); Empirical modelling; Discrete wavelet transform (DWT); Denoising
资金
- IAEA of the CRP on online monitoring systems for Research Reactors [T34001]
In this paper a detailed method for fault detection of an in-core three wires Resistance Temperature Detectors (RTD) sensor is introduced. The method is mainly based on the dependence of the fuel rod temperature profile on control rods elevation and coolant flow rate in a given nuclear reactor. For the implementation, an artificial neural network (ANN) technique has been developed to model the dynamic behaviour of the considered temperature sensor. In order to have more refined model estimation, ANN has been combined with additional noise reduction algorithms. The effective denoising work was done via the discrete wavelet transform (DWT) to remove various kinds of artefacts such as inherent measurement noise. The principle of the adopted fault detection task is based on the calculation of the difference between the ANN model estimated temperature and the online being measured temperature and then compare the deviation with a certain detection threshold to decide the sensor fault. The efficiency of the method is evaluated first on a simulated case and then on the on-line measurements obtained from a real plant. Results confirm the capacity of the developed ANN-based model to estimate a fuel rod temperature with a reasonable accuracy. (C) 2014 Elsevier Ltd. All rights reserved.
作者
我是这篇论文的作者
点击您的名字以认领此论文并将其添加到您的个人资料中。
推荐
暂无数据