Structural performance of geopolymer concrete

Abstract

Upscaling of two geopolymer concrete mixtures in an individual prestressed girder in self-compacting geopolymer concrete (SCGC) with a topping layer cast in-situ by a ready-mix geopolymer concrete provider. The objective is to study the different material properties of the mixtures, determine their impact in the structural performance and define the extent of applicability of current methods of analysis for conventional concrete structures, as defined in EN 1992-1-1 and the Rijkswaterstaat’s Guidelines for NLFEA, for geopolymer concrete. The structural performance of the elements subjected to flexural (at 28 days and 9 months) and shear (at 28 days) tests is studied by analytical methods and 2D plane stress nonlinear finite element analyses, which are compared to experimental results in terms of deformations, load-deflection response, principal strains, normal stresses, damage evolution, cracking stages, maximum load and failure mechanism; the effect of the long-term material properties (e.g. elastic modulus, creep and shrinkage) of the geopolymer concrete mixtures in the prestressing losses, cracking load and maximum load carrying capacity is analyzed. The elastic modulus of both geopolymer concrete mixtures decreases when exposed to drying and is lower than the estimates from EN 1992-1-1 for OPC concrete of the same strength class. The increase of creep and shrinkage between 30 and 60 days suggest a pronounced viscous mechanical response. The prediction models from EN 1992-1-1 for conventional concrete to determine the material properties from the 28-day compressive strength do not capture the long-term material properties. The isotropic elasticity-based prediction models underestimate considerably the creep coefficient and shrinkage strains leading to unsafe design assumptions. The short-term flexural resistance is higher (8% by analytical model and 3% for the numerical simulation) than the maximum load during testing. The stress distributions at characteristic phases and the cracking load (3% higher than the experiment) are practically equal from the analytical and numerical model. The flexural cracking pattern in the topping in the NLFEA shows more cracks at smaller spacing because higher stresses transfer from the precast girder due to perfect bond. The short-term shear resistance was 12% higher from the numerical simulation and 16% lower from the analytical model, as compared to the experiment. The cross-section of the numerical simulation is stiffer since the precast girder and the topping are perfectly bonded, in the experiment debonding occurred. The position and orientation of the shear critical crack causing failure from the NLFEA was consistent with the DIC observations. The prestress losses at 28 days are higher than for conventional concrete and increase from 26% after 28 days to 38% after 270 days. The variation in the elastic modulus at 28 days has a marginal effect in the short-term response to the flexural test. The cracking load is decreased by 15% in the specimen after 9 months but the prestress losses will continue to increase over time and the cracking resistance will decrease which is critical for the performance over the service lifetime. The design criteria of conventional concrete result in non-conservative estimates of the cracking resistance and flexural load carrying capacity.