Reinforced alkali-activated concrete beams under shear loads – a lattice model study

Abstract

The construction industry is continuously seeking alternatives to ordinary Portland cement concrete due to the high environmental impact related to cement production and global 𝐶𝑂₂ emissions. Slag-based alkali-activated concrete (AAC) has been introduced as a promising alternative due to its lower environmental footprint and its ability to achieve mechanical properties similar to conventional concrete. However, uncertainties remain regarding its long-term structural behaviour, particularly its shear behaviour. Tensile strength and stiffness have been observed to decrease over time in slag-based AAC, while compressive strength generally increases, and the combined effect of these changes on structural performance is not yet well understood.

This thesis presents the development and validation of a lattice-based numerical model for simulating the shear behaviour of slag-based AAC beams. A lattice approach was selected because it represents cracking through the failure of discrete elements, allowing crack paths to emerge naturally from local stress states without requiring the regularization techniques used in smeared continuum models. In addition, the model builds on an existing lattice framework developed by Gu et al. (2023), enabling comparison with previous studies.

The model is calibrated using the experimental dataset of Straathof (2025) for slag-based AAC-C30, which provides compressive strength, tensile strength, elastic modulus, and fracture energy at both 28 days and 6 months. These two time steps enable evaluation of both early-age and long-term behaviour using the same beam geometry. Calibration is performed through direct tensile and pull-out simulations before application to reinforced concrete beams with and without stirrups. Different mesh sizes are investigated to assess the influence of discretization on crack propagation and structural response in both slag-based AAC and ordinary Portland cement concrete beams. Data from the URBCON project are also used to evaluate the model for a different geometry and loading configuration without recalibration.

The model reproduces the experimentally observed behaviour with good accuracy. Predicted peak loads closely match the experimental values for beams with and without stirrups at both 28 days and 6 months. The results indicate that the shear capacity of slag-based AAC-C30 increases by approximately 20% between 28 days and 6 months because the increase in compressive strength outweighs the reduction in tensile strength. However, the post-peak response becomes more brittle, consistent with the dominant role of tensile strength in crack localization after formation of the diagonal shear crack. From a design perspective, the primary long-term concern for slag-based AAC is therefore the reduction in ductility and warning before failure rather than a reduction in shear capacity.

The mesh sensitivity analysis shows that finer meshes predict peak loads within 5% of the experimental results, whereas a 40 mm mesh significantly overestimates stiffness and capacity because it cannot accurately reproduce the bond-slip behaviour observed during pull-out calibration. The 15 mm mesh produces crack patterns that best match the experiments but also results in a more brittle post-peak response than the 25 mm mesh, illustrating the trade-off between crack localization accuracy and post-peak ductility in deterministic lattice models. Finally, simulations using the URBCON dataset demonstrate that the calibrated lattice model successfully captures asymmetric shear behaviour and damage localization, supporting its applicability to practical applications.

Three main limitations should be considered. First, the conclusions apply specifically to slag-based AAC and are based on a single mixture evaluated at two time steps. Second, the deterministic representation of material properties produces a more brittle post-peak response than observed experimentally. Third, the current representation of fracture energy does not fully capture gradual softening. Future research should focus on stochastic material fields and improved fracture energy modelling, particularly for shear-critical beams without stirrups.