Including model uncertainty in risk-informed decision making

Abstract

Model uncertainties can have a significant impact on decisions regarding licensing basis changes. We present a methodology to identify basic events in the risk assessment that have the potential to change the decision and are known to have significant model uncertainties. Because we work with basic event probabilities, this methodology is not appropriate for analyzing uncertainties that cause a structural change to the model, such as success criteria. We use the Risk Achievement Worth (RAW) importance measure with respect to both the core damage frequency (CDF) and the change in core damage frequency (ACDF) to identify potentially important basic events. We cross-check these with generically important model uncertainties. Then, sensitivity analysis is performed on the basic event probabilities, which are used as a proxy for the model parameters, to determine how much error in these probabilities would need to be present in order to impact the decision.A previously submitted licensing basis change is used as a case study. Analysis using the SAPHIRE program identifies 20 basic events as important, four of which have model uncertainties that have been identified in the literature as generally important. The decision is fairly insensitive to uncertainties in these basic events. In three of these cases, one would need to show that model uncertainties would lead to basic event probabilities that would be between two and four orders of magnitude larger than modeled in the risk assessment before they would become important to the decision. More detailed analysis would be required to determine whether these higher probabilities are reasonable. Methods to perform this analysis from the literature are reviewed and an example is demonstrated using the case study. We then look at policy issues surrounding the effects of uncertainty in decision making related to nuclear power generation.

Fast uncertainty quantification of activation sequences in patient‐specific cardiac electrophysiology meeting clinical time constraints

“…The key aspect for applying multifidelity methods is therefore the availability of fast low‐fidelity models. Typical examples are the following: • A mathematical model of expert opinions or empirical evidence . In this case, the challenge is to create a model that in analytic form can make use of the full stochastic input and to quantify its effect on the output. • Surrogate models, fitting data generated by the high‐fidelity model at a small number of given input samples .…”

Effectiveness of Different Model-Form Quantification Techniques as Applied to Thermal Physics-Based Simulations

“…14 Reinert and Apostolakis also used this approach in the assessment of risk for decision-making processes. 15 Multiple derivations of the adjustment factor approach exist in the literature. These derivatives all employ a similar technique of quantifying the model-form uncertainty through the use of expert opinion regarding a model's accuracy with respect to other models in the model set, or experimental data if available, by assigning model probabilities, and updating those probabilities through the use of Bayes' theorem upon the availability of experimental data.…”

“…In assuming a log normally distributed form for this factor, the first and second moments-expected value and variance-of the adjustment factor are calculated as shown in Eqs. (14)(15).…”

Representation of analysis results involving aleatory and epistemic uncertainty.

“…Such treatment is particularly important in computational analyses that are used to support important societal decisions on issues related to climate change [15][16][17][18][19], reactor safety [20][21][22][23][24][25][26], radioactive waste disposal [27][28][29][30][31][32][33][34], nuclear weapon safety [35][36][37][38], economic policy [39][40][41][42][43], environmental degradation [44][45][46][47], and many additional areas of concern and challenge. Indeed, it is difficult to envision how adequately informed decisions can be made on such issues without an appropriate assessment of the uncertainties present in the supporting analyses.…”