Transcription
Aleatory vs. Epistemic Uncertainties:Principles and ChallengesNathan Siu, Julie MarbleOffice of Nuclear Regulatory ResearchU.S. Nuclear Regulatory CommissionPresentation at ASME PVP ConferenceBaltimore, MDJuly 20, 20111
"Aleatory" and "Epistemic" UncertaintiesTerminology/concepts built into multiple documents, e.g., ASME/ANS PRA Standard Regulatory Guides– 1.2001 200– 1.174 NRC Reports– NUREG-1855– NUREG-1806– NUREG/CR-6833 Trainingaleatory uncertainty: the uncertainty inherentin a nondeterministic (stochastic, random)phenomenon is reflected by modeling thephenomenon in terms of a probabilisticmodel cannot be reduced by theaccumulation of more data or additionalinformation.epistemic uncertainty: the uncertaintyattributable to the incomplete knowledge abouta phenomenon that affects our ability to modelit is reflected in ranges of values forparameters, a range of viable models, thelevel of model detail, multiple expertinterpretations, and statistical confidence can be reduced by the accumulation ofadditional information.2
Observations Common understanding ofconceptual modeling framework inroutine Level 1 applications ChallengesCh llarisei withith ddeparturestfrom routine, e.g.,– Integrated phenomenological modeling– New issues (e.g., digital systems)– Very large uncertainties3
Challenges Operationalization of conceptual framework(characterization and incorporation ofuncertainties in computational model)– Model parameters– Model structure– Model scope/boundary conditions– Input data Communication of uncertainties and use indecision making4
Some Challenge Sources Communication precision and completeness– What is the X in P{X H}? What is H?– What is the scope of the system being analyzed?– What is the level of decomposition? Absolute vs. relative views on classification User perceptions on usefulness5
Risk Communication and Decision Support ASME/ANS RA-S-2009 HLR-QU-E:Uncertainties in the PRA results shallbe characterized. Sources of modeluncertainty and related assumptionsshall be identified, and their potentialimpact on the results understood. Decision problem may affect modeland aleatory/epistemic allocation Allocation may affect perceptionDecisionModel6
Recent and Upcoming Activities U. Stavanger Workshop "On theAssessment and Communication of Riskand Uncertainties in a Practical DecisionMaking Context: Beyond TraditionalFrameworks " January 2010Frameworks, OECD/NEA/CSNI/WGRISK: technicaldiscussion on risk communication todecision makers, April 2011 PSAM 11/ESREL 2012, Technical Sessionon risk communication to decision makers,June 20127
BACKUP8
Conceptual PRA Model - Aleatory UncertaintiesInitiating Event(IE)Safety cy1NF1BernoulliFailurePoisson2YF2SS FailsSS FailstoStartSS FailstoRunSS-FTSSS-FTR9
Conceptual PRA Model – Epistemic UncertaintiesInitiating Event(IE)Safety IEFailure2YF2SS FailsSS FailstoStartSS FailstoRunSS-FTSSS-FTRpSS-FTSpSS-FTR10
be characterized. Sources of model uncertainty and related assumptionsuncertainty and related assumptions shall be identified, and their potential impact on the results understood. Decision problem may affect model and aleatory/epistemic allocation Allocation may affect perception 6 Decision Model
