• 2019-11-10
• Opeyemi D.A, Aladegboye, O.J., Aliu, A.O
• Volume 1, Issue 1

Abstract

Information comes in various forms in scientific, technological and engineering problems. Often times the quantification methods for processing this information frequently involve unreliable assumptions and subjective judgments, and decision makers/analysts may have considerable uncertainty about the best mathematical form of the model to use. Subsequently, it may not always seem reasonable to characterise measurement imprecision by any single probability distribution. These and related issues can cause severe problems in quantitative modelling. Nonetheless, a proper quantification of uncertainty is a fundamental requirement to obtain realistic results and sensible decisions from engineering analyses. Concepts of imprecise probabilities have emerged as a sound and promising basis for analysis in this context. One of the main features of imprecise probabilities is the identification of bounds on probabilities for events of interest; the uncertainty of an event is characterised with two measure values (i.e. a lower probability and an upper probability). The distance between the probability bounds reflects the indeterminacy in model specifications expressed as imprecision of the models. The tails of the distributions are decisive for the failure probability but can only be determined and justified vaguely based on statistical data and expert knowledge, an analysis with an entire set of plausible probabilistic models and the identification of an associated upper bound for the failure probability are beneficial. This reduces the risk of wrong decisions due to unintended optimistic modelling. Implementations and applications have been reported on optimal inspection and maintenance schedule selection for corroded pipelines.

Download PDF