 |
The application of digital image processing [9] to problems in materials science and civil engineering has grown rapidly in recent years as witnessed by the emergence of conferences totally devoted to this topic [10]. While often applied at the structural level, analysis of digital images of cement-based materials at the micrometer level has provided a quantitative characterization tool for starting materials [11,12] and hydrated systems [13]. A digital image typically consists of a two-dimensional array of pixels, each assigned a greylevel value (normally ranging from 0 to 255) indicative of the strength of a measured signal [9], such as reflected light intensity or the flux of backscattered electrons in an SEM image. Such images can be easily analyzed, using pixel counting for example, to determine quantities such as phase volume fractions and surface areas as well as the sizes and other stereological properties of individual particles. Figure 1 shows a digital image of a typical ASTM Type I ordinary portland cement (OPC) [14], obtained using an SEM, in which the component clinker phases have been identified [15].
|
Although difficult to obtain experimentally, computationally, this concept can be easily extended to three dimensions. In our simulations, each pixel occupies a finite volume, typically 1 µm3, and is assigned to a specific mineralogical phase or porosity. In two or three dimensions, these digital images are easily mapped onto finite difference or finite element grids for the subsequent computation of properties such as electrical conductivity or elastic modulus [16]. Typically, these images are generated and analyzed in a static manner. It is, however, the ability to make dynamic modifications of such starting images through computation modelling that offers great potential for the simulation of microstructure development during hydration.