To develop a computerized intelligent system for CPR1000 with functions of fault diagnosis (FD) for safe operation of nuclear power plant (NPP) counts a great deal. Traditional procedures, earlier systems and classic uncertain artificial intelligence methodologies cannot apply well to some severe conditions like Three Mile Island (TMI) accident. To remedy these defectives, in this paper, the methodology of Dynamic Uncertain Causality Graph (DUCG) is brought in for diagnosis of NPP. Causal graphs of DUCG is used to symbolize logical relationships and inserted weighting factor can balance the representation of multi-valued causal relationships and avoid the cases of exponential scale-up of conditional probability table (CPT). With graph-based inference scheme, precise and compact results of FD can be computed and laid out in graphic forms. A DUCG model centering seven faults of secondary circuit in CPR1000 is constructed based on systematic analysis of Unit-I of Ningde NPP of China General Nuclear Power Group (CGNPG). Verification experiment is carried out for fault steam generator (SG) loss feed-water with sensor measurements from corresponding simulation machine NS-FSS manufactured by China Nuclear Power Simulation Technology Co, Ltd (CNPSC). Superiorities of DUCG in FD emerge by contrast with other techniques, which lay theoretical foundation for the promotion of development of CPR1000 intelligent system.
- Nuclear Engineering Division
Pilot Study of Diagnosis on CPR1000 With Dynamic Uncertain Causality Graph
- Views Icon Views
- Share Icon Share
- Search Site
Zhao, Y, Dong, C, & Zhang, Q. "Pilot Study of Diagnosis on CPR1000 With Dynamic Uncertain Causality Graph." Proceedings of the 2016 24th International Conference on Nuclear Engineering. Volume 4: Computational Fluid Dynamics (CFD) and Coupled Codes; Decontamination and Decommissioning, Radiation Protection, Shielding, and Waste Management; Workforce Development, Nuclear Education and Public Acceptance; Mitigation Strategies for Beyond Design Basis Events; Risk Management. Charlotte, North Carolina, USA. June 26–30, 2016. V004T14A009. ASME. https://doi.org/10.1115/ICONE24-60181
Download citation file: