Statistical Sensitivity Analysis of a Simple Nuclear Waste Repository Model

摘要

This report represents a preliminary step in what should be a comprehensive sensitivity analysis of the modeling of a nuclear waste repository. The purpose of the complete analysis will be to determine which modeling parameters and physical data are most important in determining key design performance criteria and then to obtain the uncertainty in the design for safety considerations. The theory for a statistical screening design methodology is developed for later use in the overall program. The theory was applied to the test case of determining the relative importance of the sensitivity of near field temperature distribution in a single-level salt repository to modeling parameters. The exact values of the sensitivities to these physical and modeling parameters were then obtained using direct methods of recalculation. The sensitivity coefficients found to be important for the sample problem were thermal loading, distance between the spent fuel canisters and their radius. Other important parameters were those related to salt properties at a point of interest in the repository. In particular, the screening methodology singled out four parameters related to the salt as having high sensitivities: salt conductivity, heat capacity, and density, and initial temperature of salt at the point of interest. These predictions of the screening theory were confirmed by direct recalculations which also reveal the time behavior of these sensitivities. It was found that the sensitivities of thermal loading, distance between the fuel canisters, their radius and salt density and heat capacity increase (in absolute value) during the 50-year period studied. The sensitivity of the initial temperature decreases (in absolute value) whereas the sensitivity of the salt conductivity is important at the time the temperature of the salt is equal to the response temperature. It was found that nonlinearity effects of the input parameters on the responses are negligible. (ERA citation 05:026991)