Fracture scaling of concrete under multiaxial compression

Abstract

The influence of specimen size on measured material properties in solid heterogeneous materials, such as concrete and rock, has been an issue of research and discussions for a few decades. A thorough understanding of the size effect phenomenon and the physical processes involved is imperative. An appreciable amount of experimental data on size effect can be found in literature, which mainly focuses on direct and in-direct tension, bending, and uniaxial compression. These data are used for developing and validating numerical material models of fracture and size effect. To date, however, only few and limited experimental data exists for size effect in the biaxial and multiaxial compressive regimes. Size effect experiments under multiaxial stress conditions require three-dimensional scaling, which are experimentally challenging. For such experiments, the hollow-cylinder geometry lends itself for providing permutations of various multi-axial states of stress around its inner-hole depending on the stress path applied to its boundaries. Under external hydrostatic stress, it allows for a gradual and stable pre-peak failure development (in case of quasibrittle materials) across the wall thickness from the inner-hole outwards. Hollow-cylinder tests are commonly used in the oil and gas industry as model experiments for perforation and wellbore stability studies. In this thesis, series of scaled hollow-cylinder tests were carried out on two model (cement- based) materials with varied maximum aggregate size. The objective of these tests was to enhance the knowledge about size effect and fracture processes in multi-axial compressive failure of quasibrittle materials. The focus was to get insight into the physical mechanisms underlying the observed size effect. In addition, the deformation behaviour and fracture characteristics were closely examined and analyzed. For this purpose, a high-pressure test cell was developed that enables testing of hollowcylinders with dimensions up to 200 mm outer-diameter and 300 mm length. The cell was equipped to accommodate smaller specimens in a size range 1:4. The set-up was supplemented with novel measuring device for monitoring the deformations taking place inside the inner-hole, both in radial and axial directions for all sizes. Impregnation experiments were performed on all tested specimens using fluorescent epoxy resin. Obtained crack patterns were examined using both optical microscopes and Environmental Scanning Electron Microscope (ESEM). Numerical analyses using the distinct element program PFC2D were conducted in order to obtain a more thorough understanding of the phenomenon and experimental results. Modelling took place through firstly, developing a synthetic material that is calibrated for its (micro-) parameters using a set of laboratory mechanical tests. Afterwards, a model was developed to simulate the hollow-cylinder test in two-dimensions. Analyses of the hollow-cylinder test and its size dependence using the simulated model material and test procedure were performed. In addition, size effect simulations were performed for uniaxial compression and Brazilian splitting tests. Size effect was observed in the strength of the hollow-cylinders with a consistent de-cease of strength with size. The experimental/numerical results and the performed analysis revealed the size effect in hollow-cylinder tests as a result of complex combination of structural factors (e.g. stress gradients), mechanical processes of failure including deformation, and material characteristics in terms of heterogeneity and fabric. Emphasizing only one factor in a model or hypothesis and neglecting others brings an error to the model, which could be significant. Observed size effect was dependent on aggregate size, being stronger for the mixture with smaller aggregate size. The onset of size effect in the experiments was observed linked to the commencement of nonlinearity in the stress-strain response. Microscopic examination of fracture processes at this stage showed small boundary cracks to exist with barely any crack interaction or propagation activities. This implies that material related factors contributing to onset of size effect should be linked to processes taking place at crack initiation, which are largely due to heterogeneity and distribution of defects. The predicted size effect according to Weibull theory described with reasonable success the obtained size effect trends near crack initiation levels.