The effect of statistical fluctuation, finite size error, and digital resolution on the phase percolation and transport properties of the NIST cementhydration model

The National Institute of Standards and Technology (NIST) cement hydration model starts with a three-dimensional digital image of cement particles, an d then uses cellular automaton rules to simulate the reaction of cement wit h water and the development of cement paste microstructure. This stochastic model uses a digital image with a certain resolution to represent the real ity of continuum shapes in a fairly small periodic unit cell. As such, it i s potentially subject to statistical fluctuation, finite size error, and th e effects of digital resolution. This paper evaluates the model in light of these potential sources of error, along with the effects of cement particl e size distribution (PSD), water/cement (w/c) ratio, and cement chemistry, by focusing on the phase percolation and transport property predictions of the model. Statistical fluctuation and finite size error are shown to be of minor importance in the model. The effect of digital resolution is signifi cant, however, but different from the case of a finite element or finite di fference model. Unlike these cases, the "best" resolution is not necessaril y the finest resolution, but must be chosen based on comparison with experi ment and various physical length scales present in cement paste. (C) 2001 E lsevier Science Ltd. All rights reserved.