Generation of Virtual 3D Concrete Microstructures Applied to Modeling Creep and Relaxation

Abstract

Nearly all nuclear power plants in the United States are operating past their intended lifetimes or are requesting lifetime extensions. Therefore, understanding changes to the concrete containment structure over time is crucial to evaluate the structure’s continued viability. Concrete materials are heterogeneous particulate composites that exhibit viscoelastic material properties, which can lead to slow deformation over time, causing stress redistribution and the potential for creep cracking. A code to generate random, 3D concrete microstructures has been developed with a novel overlap detection algorithm that reduces the computational time to generate a microstructure with a realistic volume fraction by 85\%. This code is paired with finite element analysis to predict the long-term viscoelastic properties of concrete, where data from these simulations are used to develop constitutive equations for the viscoelastic behavior of the homogenized concrete. To validate this work, the simulated creep behavior of concrete is compared to 800 days of experimental data that has been extended to 27 years of data using the Time-Temperature superposition (TTS) principal. Excellent agreement between the simulation results and experimental data is seen. Additionally, two-dimensional (2D) and three-dimensional (3D) simulations were compared to determine the impact of simulation dimensionality on predicted creep behavior. Boundary conditions, both morphological and mathematical, are investigated for their influence on concrete creep predictions. Lastly, short-term mortar stress relaxation is simulated to demonstrate the application of this framework to short term and early age creep The codes in this work are used to virtualize laboratory experiments, to obtain long-term creep data in a faster, cheaper manner.