Prediction of Cement Paste Elastic Properties via Computational Homogenization Using Digital Microstructures

This paper addresses the computational homogenization of cement paste, considering different types and sizes of representative volume elements (RVEs). Two microstructure models are examined: one generated from the simulation of hydration processes and the other based on microtomographic experimental data. A simplified three-phase virtual microstructure model is developed to enable a direct comparison between different model types. To generate this model, we have employed a multilevel analytical Mori–Tanaka approach to the preliminary homogenization of anhydrous cement grains and hydration products. Computational homogenization is performed using a fast Fourier transform (FFT)-based method. A very good agreement is observed between the results of the proposed simplified model and those of the original model. Moreover, local stress distributions at the microstructural level are analyzed. Our findings provide new insights into how cement paste's effective elastic properties vary with different volume element sizes and microstructure model types. We demonstrate that the required size of RVEs depends on the microstructure model type and highlight potential differences between results obtained for different RVE sizes. For relatively large RVE sizes, good agreement is observed between the results obtained from the simulated and experimentally based microstructure models. We also investigated the symmetry type of the determined effective elastic tensors. These findings may facilitate the multiscale modelling of cement paste and lead to a better understanding of the relationship between microstructural features and effective elastic behavior.