|
Abstract
|
Deterministic Safety Analysis and Probabilistic Safety Assessment (PSA) analyses are used to assess the Nuclear Power Plant (NPP) safety. The conventional deterministic analysis is conservative. The best estimate plus uncertainty analysis (BEPU) is increasingly being used for deterministic calculation in NPPs. The PSA methodology integrates information about the postulated accident, plant design, operating practices, component reliability and human behavior. The deterministic and probabilistic methodologies are combined by analyzing the accident sequences within design basis in the event trees of a postulated initiating event (PIE) by BEPU. The peak clad temperature (PCT) distribution provides an insight into the confidence in safety margin for an initiating event.
The paper deals with calculating the safety margin with 95% confidence and 95% probability in large break loss of coolant accident (LBLOCA). In the present study, five uncertain input parameters were selected. Uniform probability density function was assigned to the uncertain parameters in the selected range and these uncertainties are propagated using Latin Hypercube Sampling (LHS) technique. The sampled data for five parameters was randomly mixed by LHS to obtain 25 input sets. An event tree for the initiating event, LBLOCA inside containment, has been used from a VVER study on Level-1 PSA and four non-core damage (NCD) accident sequences were identified for this study. In the accident analysis the success and failure of safety systems reflected in event tree was appropriately modeled in the system thermal hydraulics code runs. The PCT was obtained for each of 25 code runs for each accident sequence. A Kolmogorov - Smirnov goodness-of-fit test carried out for PCTs indicated that they followed normal distribution for each of the accident sequences. The probability distribution of safety margin (difference between acceptable value and PCT) in each accident sequence was also obtained. The values of safety margin for the 95% confidence and 95% probability are estimated. The robustness of the system can be judged based on this. This paper describes the methodology. LBLOCA in a VVER type reactor is considered as an example.
Language: en |