Quantification of margins and uncertainties: Alternative representations of epistemic uncertainty
Introduction
In 2001, the National Nuclear Security Administration (NNSA) of the U.S. Department of Energy (DOE) in conjunction with the national security laboratories (i.e., Los Alamos National Laboratory, Lawrence Livermore National Laboratory, and Sandia National Laboratories) initiated development of a process designated Quantification of Margins and Uncertainties (QMU) for the use of risk assessment methodologies in the certification of the reliability and safety of the nation's nuclear weapons stockpile. A previous presentation describes the basic ideas that underlie QMU and illustrates these ideas with two notional examples [1]. An additional presentation illustrates the ideas underlying QMU with examples from three real analyses involving reactor safety and radioactive waste disposal that incorporate a separation of aleatory and epistemic uncertainty [2].
The analyses in Refs. [1], [2] employ probability as the mathematical structure used to represent both aleatory uncertainty and epistemic uncertainty. As a reminder, aleatory uncertainty arises from an inherent randomness in the properties or behavior of the system under study, and epistemic uncertainty derives from a lack of knowledge about the appropriate value to use for a quantity that is assumed to have a fixed value in the context of a particular analysis (Section 2 in Ref. [1]; also, Refs. [3], [4], [5], [6], [7], [8], [9], [10], [11], [12], [13], [14], [15]). Recently, several alternatives to probability for the representation of epistemic uncertainty have been introduced, including interval analysis [16–21], possibility theory [22], [23], [24], [25], [26] and evidence theory [27], [28], [29], [30], [31], [32], [33], [34].
The purpose of this presentation is to illustrate alternatives to probability for the representation of epistemic uncertainty in QMU-type analyses. The following topics are considered: the mathematical structure of alternative representations of uncertainty (Section 2), alternative representations of epistemic uncertainty in QMU analyses involving only epistemic uncertainty (Section 3), and alternative representations of epistemic uncertainty in QMU analyses involving a separation of aleatory and epistemic uncertainty (Section 4). The examples in 3 QMU with epistemic uncertainty: characterization with alternative uncertainty representations, 4 QMU with aleatory and epistemic uncertainty: characterization with alternative uncertainty representations use the same two notional examples previously used in Ref. [1]. The presentation then ends with a summary discussion (Section 5). The content of this presentation is adapted from Sections 8–10 of Ref. [35].
Of indicated representations, interval analysis and probability theory are the most widely used. The introduction of other uncertainty representations such as possibility theory and evidence theory has been accompanied by an extensive discussion of their potential use as alternatives to probability theory for the characterization of uncertainty, with some individuals maintaining that the use of these alternative representations is essential to an adequate representation of uncertainty in situations involving limited information and other individuals maintaining that probability provides the only appropriate structure for the representation of uncertainty [36], [37], [38], [39], [40], [41], [42], [43], [44]. For perspective, several comparative discussions of these different approaches to the representation of uncertainty are available [45], [46], [47], [48], [49], [50], [51], [52], [53]. The view of the authors is that interval analysis and probability theory will remain the dominant structures used in the representation of uncertainty but there will be analysis situations involving limited information where the use of other representations for uncertainty, especially evidence theory, could be appropriate and beneficial.
It is important to recognize that, at least at an intuitive level, the need for alternatives to probability theory for the representation of epistemic uncertainty has already been introduced into the discussion of QMU. In particular, the following statement appears on p. 47 of Ref. [54]: “we require positive evidence that a nuclear weapon will work; absence of evidence that it will not work is not sufficient.”. The preceding quote is an informal expression of the uncertainty information that evidence theory is intended to formally capture and communicate. As discussed in Section 2, evidence theory provides two measures of uncertainty: belief and plausibility. In the context of the preceding quote, belief provides a measure of the amount of “positive evidence” that supports the truth of a proposition, and plausibility provides a measure of the “absence of evidence” that refutes the truth of a proposition. Related concepts of necessity and possibility are present in possibility theory as discussed in Section 2. Thus, it should not be assumed that probability provides the only possible mathematical structure for the representation of uncertainty in QMU analyses. The possible use of alternatives to probability for the characterization of epistemic uncertainty is recognized in the National Academy of Science/National Research Council (NAS/NRC) report on QMU (p. 29, Ref. [55]).
Section snippets
Alternative representations of uncertainty
This section provides a brief overview of the following mathematical structures that can be used in the representation of epistemic uncertainty: interval analysis (Section 2.1), possibility theory (Section 2.2), evidence theory (Section 2.3), and probability theory (Section 2.4). The content of this section is a lightly edited reproduction of material contained in Section 2 of Refs. [56], [57]. Specifically, the content of this section does not constitute new results and is included for the
QMU with epistemic uncertainty: characterization with alternative uncertainty representations
Examples illustrating the use of the alternative uncertainty representations introduced in Section 2 are now presented for QMU problems involving only epistemic uncertainty. Specifically, the example problem in Section 3.4 of Ref. [1] is expanded to include the alternative uncertainty representations in Section 2 (Section 3.1). Then, the following topics are considered: epistemic uncertainty without a specified bound (Section 3.2), epistemic uncertainty with a specified bound (Section 3.3),
QMU with aleatory and epistemic uncertainty: characterization with alternative uncertainty representations
Examples illustrating the use of the alternative uncertainty representations introduced in Section 2 are now presented for QMU problems involving aleatory and epistemic uncertainty. Specifically, the example problem in Section 3.6 of Ref. [1] is expanded to include the alternative uncertainty representations in Section 2 (Section 4.1). Then, the following topics are considered: epistemic uncertainty in an exceedance probability deriving from aleatory uncertainty (Section 4.2), epistemic
Summary discussion
The appropriate representation of the effects and implications of uncertainty is an essential part of any analysis on which important decisions will be based. Without an adequate uncertainty representation, appropriately informed decisions cannot be made. The importance of an adequate uncertainty representation in support of appropriately informed decisions is recognized by the NNSA through its mandate for QMU (see Section 1 in Ref. [1]).
In turn, the development of an appropriate uncertainty
Acknowledgments
Work performed at Sandia National Laboratories (SNL), which is a multiprogram laboratory operated by Sandia Corporation, a Lockheed Martin Company, for the U.S. Department of Energy's National Nuclear Security Administration under Contract No. DE-AC04-94AL85000. The United States Government retains and the publisher, by accepting this article for publication, acknowledges that the United States Government retains a non-exclusive, paid-up, irrevocable, world-wide license to publish or reproduce
References (83)
- et al.
Guest editorial: treatment of aleatory and epistemic uncertainty in performance assessments for complex systems
Reliability Engineering and System Safety
(1996) Uncertainties in risk analysis: six levels of treatment
Reliability Engineering and System Safety
(1996)Uncertainty in probabilistic risk assessment
Reliability Engineering and System Safety
(1996)When to separate uncertainties and when not to separate
Reliability Engineering and System Safety
(1996)Probability, conditional probability and complementary cumulative distribution functions in performance assessment for radioactive waste disposal
Reliability Engineering and System Safety
(1996)Fuzzy sets as a basis for a theory of possibility
Fuzzy Sets and Systems
(1978)- et al.
Two views of belief: belief as generalized probability and belief as evidence
Artificial Intelligence
(1992) - et al.
Different methods are needed to propagate ignorance and variability
Reliability Engineering and System Safety
(1996) - et al.
Practical representations of incomplete probabilistic knowledge
Computational Statistics & Data Analysis
(2006) - et al.
Fuzzy sets, probability and measurement
European Journal of Operational Research
(1989)
An exploration of alternative approaches to the representation of uncertainty in model predictions
Reliability Engineering and System Safety
Uncertainty in system analysis: probabilistic versus nonprobabilistic theories
Reliability Engineering and System Safety
Fuzzy sets
Information and Control
Sensitivity analysis in conjunction with evidence theory representations of epistemic uncertainty
Reliability Engineering and System Safety
A sampling-based computational strategy for the representation of epistemic uncertainty in model predictions with evidence theory
Computational Methods in Applied Mechanics and Engineering
About assessing and evaluating uncertain inferences within the theory of evidence
Decision Support Systems
Modeling uncertainty with belief functions in numerical models
European Journal of Operational Research
Uncertainty and sensitivity analysis in the presence of stochastic and subjective uncertainty
Journal of Statistical Computation and Simulation
Characterization and evaluation of uncertainty in probabilistic risk analysis
Nuclear Safety
Propagation of uncertainty in risk assessments: the need to distinguish between uncertainty due to lack of knowledge and uncertainty due to variability
Risk Analysis
Treatment of uncertainty in performance assessments for complex systems
Risk Analysis
The concept of probability in safety assessments of technological systems
Science
Parametric uncertainty in hydrologic modeling
Transactions of the ASAE
On the quantitative definition of risk
Risk Analysis
Aleatory or epistemic? Does it matter?
Structural Safety
Applied interval analysis
Applications of interval computations
Numerical toolbox for verified computing. I. Basic numerical problems
Interval methods for systems of equations
Methods and applications of interval analysis
Interval analysis
Possibility theory: qualitative and quantitative aspects
Possibility theory part I: measure- and integral-theoretic ground-work; Part II: conditional possibility; Part III: possibilistic independence
International Journal of General Systems
Possibility theory: an approach to computerized processing of uncertainty
Belief functions and statistical inference
The Canadian Journal of Statistics
Mathematical theory of evidence
Cited by (84)
Identifying and evaluating spindle tool-tip dynamic response under different workloads
2023, Mechanical Systems and Signal ProcessingConstruction of adaptive Kriging metamodel for failure probability estimation considering the uncertainties of distribution parameters
2022, Probabilistic Engineering MechanicsA systematic review of multivariate uncertainty quantification for engineering systems
2021, CIRP Journal of Manufacturing Science and TechnologyHow to treat expert judgment? With certainty it contains uncertainty!
2020, Journal of Loss Prevention in the Process Industries