Development of empirical parameter network to enhance accident monitoring resilience

• We identify unreliable instrumentation and provide alternative signals.
• Proposed the Empirical Parameter Network based on statistically related parameters.
• Connectivity determines which parameters are most important from an I&C perspective.
• Proposed method was demonstrated through various SBLOCA scenarios.

During the Fukushima Daiichi nuclear accident, the plant operators' ability to observe the status of the power plant using the instrumentation and control (I&C) system was severely hampered by breakdowns in the plant's network, caused by the earthquake and tsunami. Thus it was difficult to obtain essential information for monitoring the internal situation of the power plant. Also, missing and incorrect information on status caused confusion, which then led to an accident. Clearly, it is crucial that I&C systems maintain the ability to monitor the internal state of reactors, even in such an inferior working environment.Herein we propose a method to identify unreliable instrumentation and to provide alternative signals. Our method, called the Empirical Parameter Network (EPN), provides estimates to replace faulty information based on statistically related parameters, and includes visualizations and other tools to enable recognition of various scenarios. The EPN included essential parameters that were selected on the basis of a literature survey, and was based on statistical analysis of an array of simulated post-accident data. The behavior of each parameter was identified and a data visualization technique was developed to intuitively display parameter correlation information. Connectivity analysis to reveal associations was performed based on the data visualization results. By incorporating the concept of connectivity, we were able to determine which parameters were most important from an I&C perspective, allowing further construction of the EPN. This newly constructed EPN will propose an alternative signal when an incorrect input signal is generated, or even when a specific parameter is missing altogether.The proposed method was demonstrated through various scenarios originating from an initial SBLOCA event, which is considered to be one of the greatest contributors to overall severe accident risk. In this research, the relationships between parameters were confirmed based on analysis of connectivity during an accident. Overall, in the damaged network condition, the reliability of the monitoring system can be improved by using the relationships between the parameters. This research can be helpful in managing accidents.