Correlated sensitivities in reliability analysis and probabilistic Burland-and-Burbidge method. Issue 3 (3rd July 2022)

• Record Type: Journal Article
• Authors: Low, Bak Kong
• Is Part Of: Georisk. Volume 16:Issue 3(2022)
• Page Start: 397
• Page End: 412
• Publication Date: 2022-07-03
• Subjects: Reliability -- FORM -- probability; Environmental risk assessment -- Periodicals; Hazardous geographic environments -- Risk assessment -- Periodicals; Reliability (Engineering) -- Periodicals; Risk management -- Periodicals; 624.151
• DOI: 10.1080/17499518.2020.1861636
• Languages: English
• ISSNs: 1749-9518

Abstract

ABSTRACT: The first-order reliability method (FORM) is regarded by most practitioners as difficult in concept and implementation, despite its merits over the deterministic approach. Another hurdle in adopting probabilistic approach is the lack of data for estimating statistical inputs. This paper first illustrates a simple reliability-based design (RBD) with respect to the rotational limit state involving two correlated random variables, to facilitate understanding of reliability analysis using FORM. It is demonstrated that RBD-via-FORM offers valuable information and insights on parametric sensitivities at its design point. An intuitive graphical perspective in the space of correlated dimensionless sensitivity indicators is explained as a complementary alternative to the perspective in the space of original random variables. The effect of parametric correlations on parametric sensitivities is also investigated. This is followed by statistical analyses of the compressibility data in the empirical Burland and Burbidge method of settlement prediction, and subsequent probabilistic analysis of the serviceability limit state (SLS) of a water storage tank which was originally designed based on Eurocode 7 (EC7). Comparisons are made between reliability analysis and EC7 design. The importance of distinguishing negative from positive reliability indices is also explained. The complementary role of RBD-via-FORM to partial factor design approach is suggested.