Back to home
An integrated data-driven methodology for early fault detection and diagnosis in nuclear power plant   WANG Hang1, PENG Minjun1, WANG Gang1, LI Wei1, LIU Yongkuo1, CHENG Shouyu1, and AYODEJI Abiodun1, 2   1. Key Subject Laboratory of Nuclear Safety and Simulation Technology, Harbin Engineering University, Harbin, Heilongjiang, 150001, China (wanghang1990312@126.com) 2. Nuclear Power Plant Development Directorate, Nigeria Atomic Energy Commission, Abuja, Nigeria (1158435404@qq.com)   Abstract: In a nuclear power plant, severe accident may result from unchecked malfunctions and human error, and it could lead to radiological release to the immediate surroundings and global ecological environment. However, a reliable fault detection and diagnosis methodology could inform the operators about the current situation of the plant and assist them to locate and diagnose the malfunction accurately and properly. This paper presents an integrated data-driven methodology for an on-line fault diagnosis in a nuclear power plant. One of the merits in this methodology is that it utilizes all the plant measured parameters to perform on-line training and simultaneously implement diagnostic task. In addition, the related algorithms in different phases of the diagnostic system are optimized to avoid incorrect results and reduce false alarms. The proposed method utilizes improved principle component analysis model to detect abnormalities. Furthermore, on-line artificial immunity algorithm is adopted to recognize the fault type based on the already existing simulation model. Consequently, some typical distance formulae for similarity measurement – the Euclidean distance and the Mahalanobis distance - are applied for on-line failure degree evaluation. The performance of this methodology is verified by applying it to the Reactor Coolant System of a Pressurized Water Reactor. The results show that this improved data-driven methodology utilized for fault detection and diagnosis is feasible and practical. More significantly, it is handy to enhance the research depth on computerized operator support system and on-line risk monitoring, which will assist operators to make decision and operation. Keyword: fault diagnosis; principle component analysis; artificial immunity algorithm; similarity measurement; pressurized water reactor
	Click here for viewing full text.