Design of comprehensive diagnosis system in nuclear power plant. (November 2017)

• Record Type: Journal Article
• Authors: Li, Wei; Peng, Min-jun; Yang, Ming; Xia, Geng-lei; Wang, Hang; Jiang, Nan; Ma, Zhan-guo
• Is Part Of: Annals of nuclear energy. Volume 109(2017:Nov.)
• Page Start: 92
• Page End: 102
• Publication Date: 2017-11
• Subjects: Diagnosis -- Sensor -- AAKR -- SPRT -- Qualitative analysis; Nuclear energy -- Periodicals; Nuclear engineering -- Periodicals; 621.4805
• DOI: 10.1016/j.anucene.2017.05.024
• Languages: English
• ISSNs: 0306-4549

Abstract

Highlights: AAKR + SPRT are applied to the condition monitoring of sensors; and to improve the interpretability of diagnosis results, a qualitative analysis method is applied to the condition monitoring of components and subsystems in a NPP. AAKR models of various sensor groups are developed and SPRT gives the online monitoring results by analyzing the residuals between the actual measurements and model predictions. Qualitative rule base, knowledge base, and fault base of Qinshan NPP are developed based on a qualitative conservation analysis method. A comprehensive diagnosis system consists of the AAKR and SPRT for sensors, qualitative analysis method for components and subsystems. In the CDS, sensor measurements after validated by AAKR and SPRT are used for the diagnosis of subsystems and components.Then the monitoring results of components and subsystem can be guaranteed to some extent. Abstract: Safe operation of a nuclear power plant (NPP) is affected by the following two main factors in addition to the human factors in practice: one is the instruments, namely the condition of the sensors; the other is the systems and components. In this paper, a comprehensive diagnosis system (CDS) for NPP is developed based on these two factors. Thus the developed CDS mainly consists of two continuous parts: firstly it is the monitoring of sensors, and then it is the diagnosis of components and subsystems in a NPP. Based on the verified measurements of sensors, the monitoring results of components and subsystems can be guaranteed. Auto-associative kernel regression (AAKR) and sequential probability ratio test (SPRT) are combined to monitor the condition of sensors. If anomaly is detected, it is to reconstruct the measurements of faulty sensors, and the verified measurements are transmitted to the following step of CDS: the diagnosis of components and subsystems. A qualitative analysis method is applied in the diagnosis of components and subsystems based on the verified sensor measurements. Simulations of fault detection and identification on the sensors and components in reactor coolant system of Qinshan NPP are carried out, and the results demonstrate the effectiveness of the proposed CDS.