A CT image-driven computational framework for investigating complex 3D fracture in mesoscale concrete

Abstract

Accurate prediction of fracture in concrete depends on the characterization of real meso-structures. Micro X-ray computed tomography (CT) can fulfil this task but is still too expensive and time-consuming to repetitively generate numerical models, also with massive degrees of freedom (DOFs) that prohibit nonlinear computations. To address such problems, we develop a novel computational framework driven by only one CT image for 3D fracture investigation of realistic mesoscale concrete models. The voxelized CT data of aggregates is first processed to obtain spatially continuous surfaces. A number of aggregates with real shapes from a created library are then randomly packed into containers through compaction and vibration simulated by a dynamic physics engine. In this way, realistic concrete models with various aggregate contents (up to 60%) can be achieved without performing extra CT tests and image processing. To capture potential micro-cracks, cohesive interface elements with zero-thickness are automatically inserted between solid elements. The proposed models with more than 15 million DOFs are well validated through typical uniaxial tension tests, elucidating complex formation mechanisms of micro-cracking, pre-peak nonlinearity, softening behaviour and 3D fracture networks. Quantitative analyses indicate that with the increase of aggregate content, the coupling effect of aggregate blocking and interfacial cracking becomes stronger to make the crack surfaces more tortuous with higher ITZ-mortar area ratios and lower load-carrying capacities, although more aggregates are often required to enhance overall elastic modulus and compressive strength, and reduce carbon footprints of concrete.