Copula-based reliability and sensitivity analysis of aging dams: Adaptive Kriging and polynomial chaos Kriging methods

• Published in: Applied soft computing Vol. 109; p. 107524
• Main Authors: Amini, A., Abdollahi, A., Hariri-Ardebili, M.A., Lall, U.
• Format: Journal Article
• Language: English
• Published: Elsevier B.V 01.09.2021
• Subjects: Dams; Global sensitivity analysis; Metamodel; Polynomial chaos Kriging; Time-dependent reliability; Uncertainty quantification; Vine copula; Dams; Polynomial chaos Kriging; Vine copula; Uncertainty quantification; Global sensitivity analysis; Time-dependent reliability; Metamodel
• ISSN: 1568-4946; 1872-9681
• DOI: 10.1016/j.asoc.2021.107524

Time-dependent reliability and sensitivity analysis, in which the nature of demand, capacity and the limit state function varies over the life cycle of the structural system, is a challenging task. Meta-models are a suitable tool to construct inexpensive-to-evaluate and accurate surrogates in analysis of modern engineering problems. The paper’s contribution is threefold: First, the superiority of polynomial chaos Kriging (PCK) meta-model in terms of efficiency and accuracy is depicted in comparison to two other state-of-the-art methods in an explicit representation of a pilot gravity dam. Second, the problem of aging dams is analytically studied. Adaptive reliability approaches which benefit from Kriging and PCK meta-models are also investigated in a comparative analysis with classical reliability methods. Third, the importance of considering nonlinear dependency between random variables by copula theory is investigated under the reliability and sensitivity concepts. The results show that PCK meta-model can be used as an effective technique in uncertainty quantification (UQ) of dams. Furthermore, Kriging and PCK-assisted reliability methods can establish fairly accurate meta-models to perform reliability analysis in structural UQ with low computational efforts. Finally, the concept of UQ is propagated to ”dam class” in which the dam shape and its age are assumed to be variable. A generalized program is developed to assist dam owners and decision-makers in approximate failure probability estimation of gravity dams. This paper paves the road for global risk assessment of dams. •Vine-copula-based reliability and sensitivity analysis of gravity dams are presented.•Nonlinear dependency among random variables with and without vine-copula is examined.•Polynomial chaos kriging is introduced as an effective meta-model in dam engineering.•Surrogate-assisted uncertainty quantification of aging dams is provided.•A software is developed for rapid failure evaluation of different aging dam classes.