A Monte Carlo method for estimation of the probability of combustion-induced containment failure

The paper will present a probabilistic approach to assessment of the likelihood of containment failure due to combustion of hydrogen and carbon monoxide using a Monte Carlo method. The analyses are performed within an event tree framework that considers three phases: (i) Phase 1 - extends from start of the accident (or alternatively, from the onset of rapid Zircaloy oxidation) up until the time of reactor coolant system (RCS) or reactor pressure vessel (RPV) breach (period prior to RCS/RPV breach); (ii) Phase 2 - corresponds to RCS/RPV breach period; and (iii) Phase 3 - corresponds to the period following Phase 2 and Molten Core Concrete Interaction (MCCI) and during the period that containment conditions would be conducive (i.e., not steam-inerted) to combustion. For each phase, the likelihood of combustion-induced containment failure is decomposed in order to enable a logical and consistent quantification. In this treatment, the conditional probability of containment failure in each subsequent phase is directly dependent on the existence of ignition sources and flammability in the prior phase (or phases). Therefore, for those event tree paths on which prior combustion events have taken place, a substantially smaller hydrogen mass will be available for a later combustion (note that the combustion of CO is only applicable to Phase 3). On the other hands, for event tree paths that involved no prior combustion events, a larger quantity of hydrogen could be available for later combustion. Table 1 lists a summary of estimated conditional containment failure probabilities for Westinghouse PWRs with large-dry containment, during various phases of severe accidents. It is seen that the overall probability of early and very early containment failure is estimated to be essentially zero for all Westinghouse plants with large-dry containment. On the other hand, the likelihood of late containment failure due to combustion is estimated to range from negligibly low values for sub-atmospheric designs to as high as about 0.1 for some of the atmospheric plants. (authors)