Do we have enough data? Robust reliability via uncertainty quantification

Abstract

A generalised probabilistic framework is proposed for reliability assessment and uncertainty quantification under a lack of data. The developed computational tool allows the effect of epistemic uncertainty to be quantified and has been applied to assess the reliability of an electronic circuit and a power transmission network. The strength and weakness of the proposed approach are illustrated by comparison to traditional probabilistic approaches. In the presence of both aleatory and epistemic uncertainty, classic probabilistic approaches may lead to misleading conclusions and a false sense of confidence which may not fully represent the quality of the available information. In contrast, generalised probabilistic approaches are versatile and powerful when linked to a computational tool that permits their applicability to realistic engineering problems.

A post-contingency power flow emulator for generalized probabilistic risks assessment of power grids

“…The epistemic uncertainty can be reduced by further data collection, at least in principle. From an intuitive standpoint, input/output of analysis for which information is abundant and high in quality will be affected by a mild or negligible effect of epistemic uncertainty (although this is generally difficult to judge a priori [13]). In those cases, pure probabilistic methods will be suitable to perform uncertainty quantification tasks (e.g.…”

“…On the other hand, in situations for which data is not abundant, the background knowledge limited and inherently variable factors are affecting the analysis, both aleatory and epistemic uncertainties will have a relevant role. In those cases, the use of pure probabilistic approaches may lead to an underestimation of the uncertainty [13]. The underestimation is attributable to modeling assumptions needed to characterize the imprecision in a probabilistic framework (e.g.…”

“…and are used in this work to model factors affected by mixed aleatory-epistemic uncertainties. P-boxes have been recently applied for reliability analysis [15], [16], [17], in the context of evidential networks [22], and also for power grid analysis [13], [14]. A P-box is defined as a pair of CDFs bounding the probability between upper and lower bounds.…”

“…Probability boxes (P-boxes) are a powerful and versatile tool and are used within the framework to model factors affected by mixed sources of uncertainty, see e.g. [12,13,14,15,16,17,11]. Uncertainty propagation and sensitivity analysis quantify the effect of imprecision and inherent variability on the risk index.…”

An efficient computational strategy for robust maintenance scheduling: Application to corroded pipelines

“…Fortunately in many engineering applications the response of the model is monotonic with respect to the imprecision of the input parameters. In general, this allows to estimate the bounds of the probability of failure with only 2 full probabilistic analysis (Rocchetta et al 2018). Advanced Line Sampling (de Angelis et al 2015) method can further reduce the computational cost allowing the estimation of the bonds of the probability of failure with only 1 efficient probabilistic analysis (de Angelis et al 2014).…”