The effect of correlations among random member properties on structural reliability

Autor: Dimos C. Charmpis
Rok vydání: 2019
Předmět:
Zdroj: Structures. 19:463-475
ISSN: 2352-0124
Popis: This paper investigates the influence that correlations among random member properties have on structural reliability analysis results. The overall input uncertainty for a reliability analysis is jointly described with a multivariate probability distribution; each marginal of the joint distribution describes the randomness of a structural property for a member or group of members of a structure, while the correlation matrix specifies the degrees of interdependence among the random member properties. This type of joint uncertainty is effectively handled with a Monte Carlo (MC) simulation procedure, which is used to compute probabilities of partial/full collapse (failure probabilities) of elastoplastic structures without restrictions on how the marginal member properties are distributed and correlated. A specialized two-step heuristic sampling approach is employed to generate multivariate random samples and induce arbitrary intercorrelations for the MC simulations: first a univariate random sample from each marginal input distribution is independently generated; then a simulated annealing optimizer rearranges the univariate random samples, in order to induce the pre-specified correlations between them. The numerical results obtained for two elastoplastic steel frames reveal significant variations of structural reliability for various correlation values and patterns of random member properties.
Databáze: OpenAIRE