A study on reliability modeling of the safety-critical network communication in NPP

Abstract

The objective of this study is development of the framework to estimates network failure probability by quantification of each path for the hazardous states. The paths which might lead a system to an unsafe state were identified using state transition diagram based on the protocol specification. Each hazardous state was also identified based on the cause of the failure and the effect on the system. Each path was quantified in a probabilistic manner, by applying appropriate data. There are four hazard states: failure of token passing, long term down, failure of sending data frame, failure of receiving data frame. Several parameters were evaluated to quantify each path. Those parameters were related to the hardware failures, software failures, and transient failures. Quantification scheme for each failure is as follows. For hardware failures, it is common to use the sum of each failure data. In this study, the demand failure probabilities of each sending/receiving demand were evaluated. For software failures, the input profile for the reliability test was suggested. For transient failures Markov model was developed. The sensitivity study shows that the software is the strongest parameter among other failure causes. The bit error rate also strongly affects the ring stability, and the hazard state probability, consequently. By expanding this study, in combination with the plant characteristics we can reflect the effects of effects of network failure in nuclear power plant to the conventional PSA model.