Modeling of complex microcracking in cement based materials by combining numerical simulations based on a phase-field method and experimental 3D imaging

An approach combining numerical simulations and experimental techniques is developed to model complex microcracking in heterogeneous cementitious materials. The proposed numerical model allowed us to predict accurately in 3D the initiation and the propagation of microcracks at the scale of the actual microstructure of a real sample subjected to compression. Its predictions have been validated by a direct comparison with the actual crack network characterized by 3D imaging techniques. In a first part, the numerical simulation tools are developed and tested. More specifically, the phase-field method is applied to microcracking simulations in highly heterogeneous microstructures and its advantages for such simulations are discussed. Then, the technique is extended to account for interfacial cracking, possibly occurring at inclusion/matrix interfaces. In a second part, the experimental methods used and developed in this work are described. The procedures to obtain the evolution of the 3D crack network within the samples by means of X-rays computed microtomography and in-situ mechanical testing are presented. Then, we focus on the developed image processing tools based on digital volume correlation to extract with good accuracy the cracks from the grey level images. In a third part, we compare the predictions of the numerical model with experimental results obtained, first, with a model material made of expanded polystyrene beads embedded in a plaster matrix, and second, to a more complex lightweight concrete. More precisely, we use the experimental data to identify by inverse approaches the local microstructural parameters, and use the experimental displacements measured by digital volume correlation to define boundary conditions to be applied on sub-domains within the sample for the simulations. The obtained direct comparisons of 3D microcrack networks and their evolutions demonstrate the very good predictive capability of the numerical model