LAUSR

Abstract In order to provide an efficient way to describe the probabilistic characteristics of chloride diffusion coefficient, a probabilistic prediction model which involves both aleatory and epistemic uncertainties was developed for chloride diffusion coefficient of concrete in terms of material parameters. The influences of important material parameters including water-to-binder ratio as well as contents of fly ash and slag on chloride diffusion coefficient of concrete were investigated based on 295 sets of experimental data measured by the rapid chloride migration method. Then a deterministic prediction model for chloride diffusion coefficient of concrete was developed based on the stepwise regression method by taking into account the influences of the above material parameters. Subsequently, a probabilistic prediction model which involves both aleatory and epistemic uncertainties was developed for chloride diffusion coefficient of concrete based on the Bayesian theory and the Markov Chain Monte Carlo method (MCMC) method. Finally, the accuracy of the proposed probabilistic prediction model was validated by comparing with experimental data and traditional deterministic prediction models. Analysis results show that the proposed probabilistic prediction model not only provides a rational approach to describe the probabilistic characteristics of chloride diffusion coefficient by means of the probability density function and the cumulative distribution function, but also provides an efficient way to calibrate the accuracy of traditional deterministic prediction models based on the confidence interval and the confidence level.