Argonne National Laboratory Reports

Techniques for probabilistic response-surface analysis have been developed to obtain the probability distributions of the consequences of postulated nuclear-reactor accidents. The uncertainties of the consequences are caused by the variability of the system and model input parameters used in the accident analysis. Probability distributions are assigned to the input parameters, and parameter values are systematically chosen from these distributions. These input parameters are then used in deterministic consequence analyses performed by mechanistic accident-analysis codes. The results of these deterministic consequence analyses are used to generate the coefficients for analytical functions that approximate the consequences in terms of the selected input parameters. These approximating functions are used to generate the probability distributions of the consequences with random sampling being used to obtain values for the accident parameters from their distributions. A computer code PROSA has been developed for implementing the probabilistic response-surface technique. Special features of the code generate or treat sensitivities, statistical moments of the input and output variables, region-wise response surfaces, correlated input parameters, and conditional distributions. The code can also be used for calculating important distributions of the input parameters. The use of the code is illustrated in conjunction with the fast-running accident-analysis code SACO to provide probability studies of LMFBR hypothetical core-disruptive accidents. However, the methods and the programming are general and not limited to such applications