 |
Although often taken for granted as a commodity material, portland cement is a complex heterogeneous particulate material. When mixed with water, a variety of reactions transform the initial suspension into a rigid, load-bearing matrix which comprises the binder phase of a typical concrete. To this date, precise knowledge of the mechanisms, stoichiometries, thermodynamics, and kinetics of the hydration reactions remains to be provided. With this state of affairs, it is often difficult to quantitatively relate the microstructure of cement paste to its ultimate properties such as strength, diffusivity, and permeability, so that material performance may be improved. The ability to accurately predict performance will play a key role in the ongoing paradigm shift from prescriptive to performance-based standards [1].
In recent years, computer modelling has been successfully applied to elucidating microstructure-property relationships of cement-based materials [2]. Such elucidation requires a two-step process: generation of a representative microstructure in the computer, and computation of the property of interest, often using finite difference or finite element techniques. Although much information has been obtained using a three-dimensional microstructure model based solely on the hydration of the predominant phase present in portland cement, tricalcium silicate (C3S 1) [ 3,4], many problems of interest to cement researchers require a microstructural representation which includes all of the major phases of portland cement. Thus, recent efforts have focused on developing a three-dimensional cement hydration and microstructure program which accounts for the multi-size and multi-phase nature of cement grains.
Computationally, this requires acquisition or generation of a representative three-dimensional starting microstructure for use as input into the hydration and microstructure development program. Here, computational techniques are developed for converting a set of two-dimensional scanning electron microscopy (SEM) images into a three-dimensional representation of a given cement. The developed procedures reproduce the particle size distribution of the cement as well as the individual phase volume and surface area fractions. The final 3-D image is then used as input for the cellular automata-based hydration model. In addition to providing a 3-D map of the microstructure as it evolves, the hydration code also outputs the degree of hydration, the heat released, and the chemical shrinkage as a function of the number of hydration "cycles" which have been executed. These three model measures can be compared against their experimental counterparts to calibrate and validate the kinetics of the cement hydration model. The model and experimental program are summarized in the flow diagram in Figure 1.
|