Sensitivity Analysis Methods for Systems with Epistemic Uncertainties

摘要

This paper examines a number of sensitivity analysis methods which can be used to determine the comparative importance of input epistemic uncertainties on the system output. These results would then indicate which input uncertainties would most merit additional study. The methods examined in this paper are local sensitivity analysis by finite difference, scatter plot analysis, variance based analysis, and p-box based analysis. The effectiveness of these methods in problems with epistemic uncertainty is compared using an example. This information is used to determine which method should be used based on the linearity of the system and the number of epistemic variables being examined.