Modeling of hydration reactions using neural networks to predict the average properties of cement paste

This paper presents a hydration model that describes the evolution of cement paste microstructure as a function of the changing composition of the hydration products. The hydration model extends an earlier version by considering the reduction in the hydration rate that occurs due to the reduction of free water and the reduction of the interfacial area of contact between the free water and the hydration products. The BP Neural Network method is used to determine the coefficients of the model. Using the proposed model, this paper predicts the following properties of hardening cement paste: the degree of hydration, the rate of heat evolution, the relative humidity and the total porosity. The agreement between simulation and experimental results proves that the new model is quite effective and potentially useful as a component within larger-scale models designed to predict the performance of concrete structures.