Calculating exceedance probabilities using a distributionally robust method. (July 2017)

• Record Type: Journal Article
• Authors: Faridafshin, Farzad; Grechuk, Bogdan; Naess, Arvid
• Is Part Of: Structural safety. Volume 67(2017)
• Page Start: 132
• Page End: 141
• Publication Date: 2017-07
• Subjects: Exceedance probability -- Probability bound -- Distributionally robust optimization -- Log-concave distribution -- Imprecise probability; Structural stability -- Periodicals; Safety factor in engineering -- Periodicals; Reliability (Engineering) -- Periodicals; Constructions -- Stabilité -- Périodiques; Coefficient de sécurité en ingénierie -- Périodiques; Fiabilité -- Périodiques; 620.86
• DOI: 10.1016/j.strusafe.2017.02.003
• Languages: English
• ISSNs: 0167-4730

Abstract

Highlights: A tight upper bound of exceedance probability is calculated. Major commonly used probability distributions are covered. First two moments of a dataset and log-concavity of the CDF are used as a basis. The methodology might be used in situations with limited amounts of data. Abstract: Calculation of exceedance probabilities or the inverse problem of finding the level corresponding to a given exceedance probability occurs in many practical applications. For instance, it is often of interest in offshore engineering to evaluate the wind, wave, current, and sea ice properties with annual exceedance probabilities of, e.g., 10 −1, 10 −2, and 10 −3, or so-called 10-year, 100-year, and 1000-year values. A methodology is provided in this article to calculate a tight upper bound of the exceedance probability, given any probability distribution from a wide range of commonly used distributions. The approach is based on a generalization of the Chebyshev inequality for the class of distributions with a logarithmically concave cumulative distribution function, and has the potential to relieve the often-debated exercise of determining an appropriate probability distribution function based on limited data, particularly in terms of tail behavior. Two numerical examples are provided for illustration.