A multiscale methods for prediction of the elastic properties of portland cement blended with fly ash based on X-CT and hydration model

This study proposed a multiscale framework for the prediction of elastic properties of Fly Ash Cement paste (FAC). The microstructure of FAC was characterized by the X-ray Computed Tomography technology (X-CT), then the phase properties of microstructure of FAC were mapped onto the equivalent macroscopic elastic modulus (E) by the asymptotic expansion homogenization method. Efficient numerical simulations supported the rapid generation of large amounts of mapping data, and Artificial Neural Network (ANN) was applied to learn the above mappings containing geometrical information of microstructure of FAC. The optimal parameter combination for the phase properties was derived from the trained ANN. The calculated E based the optimal parameter combination agreed well with the experiment, indicating the feasibility of the proposed multiscale framework. Based on the validated multiscale framework, a parametric analysis was carried out using a hydration model, and it also showed good prediction results in E of FAC with different water to binder mass ratio and fly ash (FA) content. The proposed multiscale framework provides a new insight for fast determination of microstructural properties of cementitious materials, which will contribute to accelerate the development of new materials.