Computing with Epistemic Uncertainty.

摘要

This report proposes the use of uncertainty management principles for processing combinations of aleatory and epistemic uncertainty forms through arithmetic operations to yield an output uncertainty margin as an extreme value interval. The method is demonstrated using some test equations from an Epistemic Uncertainty Workshop conducted in 2002 by the Sandia National Laboratories in the USA. The approach is robust, computationally efficient and does not require special assumptions. A comparison with the results of the workshop participants showed similar results to most workshop participants. The benefit of this approach is that it is a step-wise process for computation with no need for a large number of simulations and complex sampling strategies.