Integration of the functional reliability of two passive safety systems to mitigate a SBLOCA+BO in a CAREM-like reactor PSA

Abstract This paper describes a case study of a methodological approach for assessing the functional reliability of passive safety systems (PSS) and its treatment within a probabilistic safety assessment (PSA). The functional unreliability (FU) can be understood as the failure probability of PSS to fulfill its mission due to the impairment of the related passive safety function. The safety function accomplishment is characterized and quantified by a performance indicator (PI), which is a measure of how far the system is from verifying its mission. PI uncertainties are estimated from uncertainty propagation of selected parameters. A methodology based on the reliability methodology for passive system (RMPS) one is used to estimate the FU associated to the isolation condensers (ICs) in combination with the accumulators (medium pressure injection system) of a CAREM-like integral advanced reactor. A small break loss of coolant accident with black-out is selected as an evaluation case. This implies success of reactor shut-down (inherent) and failure of residual heat removal by active systems. The safety function to accomplish is to refill the reactor pressure vessel (RPV) in order to avoid core damage. For this case, to allow the discharge of accumulators into RPV, the pressure must be reduced by the IC. The methodology for passive safety function assessment considers uncertainties in code parameters, besides uncertainties in engineering parameters (design, construction, operation and maintenance), in order to perform Monte Carlo simulations based on best estimate (B-E) plant model. Then, response surfaces based on PI are used for improving the FU calculation precision (also exploiting they fast-run feature). Their prediction capabilities are enhanced near the failure zone by B-E simulations feedback. Finally, for FU integration into traditional PSA, two solutions for sequence modeling by means of standard (static) event trees are considered. The advantages/drawbacks of each solution are discussed afterward.