STATISTICAL LINEAR REGRESSION ANALYSIS OF PREDICTION MODELS FOR CREEP AND SHRINKAGE

ABSTRACT Several models for the prediction of creep and shrinkage of concrete are compared statistically with test data available in the literature. The models are algebraically transformed into a linearized form and statistical regression is then carried out. Although the BP Model performs distinctly better than the ACI and CEB-FIP Models, the scatter is large for all models, due to the difficulty in predicting material parameter values without any tests of the given concrete. Introduc t ion Practical probabilistic analysis of concrete creep and shrinkage has recently been rendered meaningful by extraction of an extensive body of statistical information from the literature; see Ref. i and 2, in which test data for 80 different concretes from various laboratories throughout the world, consisting of over 800 experimental curves and over i0,000 data points, have been analyzed statistically. It has been shown that, if no measurements for a given concrete are made, the uncertainty of its creep and shrinkage prediction on the basis of the chosen concrete mix parameters and the chosen design strength is enormous (and is much larger than for strength). From all the data considered in Ref. i, it was determined that, on the whole, the prediction errors that are exceeded with a 10% probability (90% confidence limits) are, for the best known practical prediction models, as follows: BP The first of these models is to a greater extent than the others based on physical considerations and is applicable over a much wider range of conditions. However, it is relatively complicated. The second model is much simpler, and still represents probably the optimum prediction method at its degree of simplicity. The desire for simplicity should not be exaggerated, however. Since the errors in the prediction of creep effects in structures are caused more by errors in prediction of creep properties than by simplifications of structural analysis, the designer should spend more time on careful 86