LAUSR

Abstract Because concrete structures are sensitive to creep and shrinkage, it is generally recognized that the probabilistic prediction results of long-term deflection might exhibit significant scatter. If the short-term measurements are available, Bayesian statistics can significantly reduce the uncertainty of prediction, thereby enabling more objective probabilistic results. In the present study, a new computational framework for Bayesian inference regarding the long-term deflection of concrete structures is proposed. Importance-sampling technology is introduced to ensure that the realization points fall within the range of the observed data as frequently as possible. Moreover, the response surface approach, which requires only several runs of finite-element analyses in stochastic analysis, is also adopted to improve computational efficiency. In addition, a stochastic process, rather than a random variable, is presented to describe the random properties of creep. Through a numerical example study, the accuracy and efficiency of the proposed improvement and modification are demonstrated.