ABSTRACT
This paper summarizes the results of the Nordic reference study on uncertainty and sensitivity analysis. Experiences from a previous Benchmark Exercise and from reference studies concerning treatment of common cause failures (CCFs) and human interactions have been used extensively. The work has been focused on generation of an uncertainty distribution for a selected accident sequence using state-of-the-art approaches to these problem areas. In addition, a comparative study of Monte Carlo codes for uncertainty propagation, practically available to the participants, has been carried out. Generally, the best estimate results obtained by different groups show good agreement. Specific findings concern e.g. consensus methods for quantification of CCF-contributions, approach to modeling of phased-mission operation, impact of application of different distributions and distribution approximations, impact of first moment approximation, importance of "state-of knowledge" dependences, approach to ranking of principal uncertainty contributors, experiences gained from applications of the Bayesian approach and assessment of the impact of a possible design modification.
