Abstract:

A multi-source large-sample model for chloride diffusion coefficient of concrete was proposed by introducing cement type factor, considering the influence of cement category and strength grade. Firstly, the database with 179 sets of rapid chloride migration (RCM) test data from 70 laboratories was employed for determination of function of chloride diffusion coefficient on water-binder ratio, the type and strength grade of cement through regression analysis. Furthermore, the cement type factor was determined while the computational model was developed for chloride diffusion coefficient of concrete by means of the two-phase regression method. Then, comparison was implemented with different computational models and experimental data, the results show that the multi-source large-sample model has a 15.7~19.6% higher fitting degree to the experimental data than the traditional mono-source small-sample model. Moreover, the cement type factor can reasonably reflect the influence of cement category and strength. By comparing with traditional models and test data, the proposed model is validated with higher predicting accuracy and wider adaptability, reducing the weighted average error by 32.0% and the coefficient of variation by 25.0%.

Key words: concrete; cement type factor, chloride diffusion coefficient, rapid chloride migration, multi-source and large-sample