Comparison of two uncertainty analysis methods

A demonstration of two methods of uncertainty analysis was carried out to assess their utility for future use in treating computer models of nuclear power systems. The two methods of uncertainty anaylsis, called the response surface method and the crude Monte Carlo method, produced comparable results for the probability density function of the peak cladding temperature as computed by a simplified nuclear code that was subjected to seven uncertainty parameters. From these density functions, the upper cumulative tail probabilities were obtained and were shown to be measures of parameter margin. The response surface method provides sensitivity coefficients and also an inexpensive framework for evaluating the effects of the various assumptions inherent in the method. The crude Monte Carlo method provides no sensitivity coefficients and requires a complete rerun if a single uncertainty input density should be changed. The response surface method is recommended for use, where economically feasible, since the advantages of the method far outweigh the disadvantages.