Uncertainty and Sensitivity Analysis for Long-Running Computer Codes: A Critical Review

Report Date: January 2010
Appendices: No

Executive Summary

This report presents a critical review of existing methods for performing probabilistic uncertainty and sensitivity analysis for complex, computationally expensive simulation models. In the context of Probabilistic Risk Assessment (PRA), these models may be used to (i) estimate the reliability of passive systems in the absence of operational data, (ii) inform Level 1 accident sequence development and event tree structure, (iii) establish Level 1 PRA success criteria, (iv) develop Level 2 PRA event tree structure and split fraction values, (v) performing Level 3 PRA offsite consequence analysis, and (vi) provide simulation capacity in dynamic PRA tools. As discussed in Chapter I, UA can be regarded as an attempt to determine what effect the inputs and their uncertainties have on the model output, while SA is an attempt to determine how these inputs and uncertainties affect the model’s output. In other words, UA is focused on propagating the input uncertainties through the model, while SA studies the relationship between the inputs and the outputs...

Program:     NSP Nuclear Systems Enhanced Performance
Type:     TR 

RPT. No.: 24