Uncertainty Due to Parameter Randomness via Sampling of Deterministic Solutions

Abstract

Concrete creep and shrinkage are notorious for high random scatter of input parameters and high uncertainty of long-time predictions. Obviously, the structural design should not be based merely on mean predictions. In this chapter, we show how to estimate realistic confidence limits on the long-time creep and shrinkage predictions. Except for creep buckling of columns and shells, these limits are, fortunately, far less stringent than those required for failure prevention, but nevertheless important to make premature structural repair or demolition rare enough. Because sustainable design calls for structural lifetimes in excess of a century, we focus on long-term predictions of structural performance in the light of the randomness of material and environmental parameters. We emphasize the numerical sampling approach based on repeated runs of deterministic analysis of creep and shrinkage effects for judiciously selected random samples of input data. Finally, our discussion is focused on improving long-term predictions by means of the Bayesian updating in which the uncertainty due to the experimental data and prediction model is reduced by prior data on short-time creep of the given concrete and on the observed initial deformations of the given structure.