Sampling analysis of concrete structures for creep and shrinkage with correlated random material parameters

Yunping Xi*, Zdeněk P. Bažant

*Corresponding author for this work

Research output: Contribution to journalArticlepeer-review

13 Scopus citations


The latin hypercube sampling method, which represents the most efficient way to determine the statistics of the creep and shrinkage response of structures, has previously been developed and used under the assumption that the random parameters of the creep and shrinkage prediction model are mutually independent. In reality they are correlated. On the basis of existing data, this paper establishes, by means of the method of maximum likelihood, the joint multivariate probability distribution of the random parameters involved, tests the hypothesis of mutual dependence of parameters on the basis of the χ2-distribution, and generalizes the latin hypercube sampling method to the case of correlated multinormal random parameters. The generalization is accomplished by an orthogonal matrix transformation of the random parameters based on the eigenvectors of the inverse of the covariance matrix. This yields a set of new random parameters which are uncorrelated (independent) and can be subjected to the ordinary latin hypercube sampling, with samples of equal probabilities. Numerical examples of statistical prediction of creep and shrinkage effects in structures confirm the practical feasibility of the method and reveal a good agreement with the scatter observed in some previous experiments.

Original languageEnglish (US)
Pages (from-to)174-186
Number of pages13
JournalProbabilistic Engineering Mechanics
Issue number4
StatePublished - Dec 1989

ASJC Scopus subject areas

  • Statistical and Nonlinear Physics
  • Civil and Structural Engineering
  • Nuclear Energy and Engineering
  • Aerospace Engineering
  • Condensed Matter Physics
  • Ocean Engineering
  • Mechanical Engineering


Dive into the research topics of 'Sampling analysis of concrete structures for creep and shrinkage with correlated random material parameters'. Together they form a unique fingerprint.

Cite this