STATISTICAL EXTRAPOLATION OF SHRINKAGE DATA - PART II: BAYESIAN UPDATING.

The statistical information previously available in the literature on concrete shrinkage is exploited to improve the extrapolation of short-time shrinkage measurements into long-time measurements. With this approach the long-time predictions can be significantly improved, compared to the predictions obtained by statistical regression from the measured data alone, as described in Part I of this study. The method of analysis consists of latin hypercube sampling of random parameters of the shrinkage formula for the prior, and in adjustments of the weights of the samples on the basis of the Bayes theorem. The shrinkage formula of the BP model, which is justified by diffusion theory, is used. An improvement of the formula for predicting the final shrinkage strain from concrete composition is also presented.