Next: 3-D Reconstruction Up: Main Previous: Introduction


2. Imaging of Cement particles

To adequately model the hydration behavior of a cement, it is necessary to accurately characterize the starting material, the cement powder. The application of scanning electron microscopy (SEM) and image analysis techniques to the characterization of cementitious materials is a relatively recent development. Scrivener [13] utilized SEM and X-ray microanalysis to determine the distribution of silicates and interstitial phases in cement grains and compared the results to the rates of heat release during hydration. Stutzman [14] identified the four major phases present in cement clinker, also utilizing a combination of SEM and X-ray microanalysis techniques. Bonen and Diamond [15] applied these techniques to analyzing the same cement ground in both a ball mill and a high pressure roller mill, by classifying the predominant phase found in each cement particle and characterizing each 2-D particle on the basis of aspect ratio and shape factor.

The combination of SEM backscattered electron (BSE) and X-ray images is also ideal for obtaining 2-D images of cement particles in which each pixel in the image has been identified as a single major cement clinker phase (tricalcium silicate, dicalcium silicate, tricalcium aluminate, tetracalcium aluminoferrite, gypsum, free lime, and the alkali sulfates) [16]. Because the BSE and X-ray images can be obtained at exactly the same locations on a sample, the information in the images can be combined and interpreted to segment the image into its constituent phases. For this procedure, the cement of interest is blended with an epoxy resin (with a viscosity similar to that of water) to form an almost dry paste. The paste is pressed into a sample mold and cured at 60 C for 24 hours. The cured specimen is then cut using a low-speed diamond wafering saw, cutting first a layer from the outer surface and then about 10 mm into the sample. This two-cut procedure produces parallel faces, minimizing the need to refocus as one traverses a specimen. The sample is appropriately polished (HTML reference here) and then coated with carbon to provide a conductive surface for viewing in the SEM.

For the SEM analysis of these types of specimens, typical operating conditions are an accelerating voltage of 12 kV with probe currents of about 2 nA and 10 nA for collecting the BSE and X-ray images respectively. In the BSE images, brightness is proportional to the average atomic number (Z) of a phase, as indicated in Table (HTML reference to Paul's chapter). The X-ray images are acquired by slowly scanning the specimen while measuring the X-ray counts within energy windows encompassing a peak for the element of interest. For this work, X-ray images are typically acquired for the following elements: Ca, Si, Al, Fe, S, and K. A scan time of about three hours is typical for acquiring a set of 512x400 X-ray images. Figures 1 and 2 show backscattered electron and Ca X-ray images for a typical Type I portland cement.

photo here


Figure 1. Backscattered electron image of a typical Type I Portland cement.

photo here


Figure 2. X-ray image for Ca for the cement in figure 1.

To perform phase segmentation based on a set of these images, the first step is to create an RGB image in which the following X-ray signal to color assignments are made: Ca to red, Si to green, and Al to blue. Figure 3 shows the resultant image obtained for the example cement. One can clearly see two shades of yellow in this image indicating the tricalcium silicate and dicalcium silicate phases. The purple regions indicate the presence of the aluminates phases, while bright and dark red represent free lime and gypsum respectively. This image is combined with the X-ray images for Fe, S, and K, to assign each pixel to be a single phase of the cement. For example, the presence of Ca, Fe, and Al indicates the tetracalcium aluminoferrite phase.

photo here


Figure 3. RGB encoded image for the cement in figure 1. Ca signal is stored in the red channel, Si in the green, and Al in the blue, so that each clinker phase may be identified on the basis of color.

The image produced from this initial segmentation, as shown in Figure 4, still contains a fair amount of noise. To sharpen the phase distinction and eliminate this noise, a median filter is applied. Here, each "solid" (non-porosity) pixel in the image is reassigned to be the phase occupied by the majority of its neighbors, excluding porosity, if this majority fraction exceeds a preset limit. Typically, the eight neighboring pixels in a 3x3 box centered at the pixel of interest are used for this filtering. Applying this filter to the image in Figure 4 results in the final segmented image illustrated in Figure 5. The final image for a second cement is illustrated in Figure 6.

photo here


Figure 4. Image after initial segmentation and phase assignment.


photo here


Figure 5. Final processed image. Tricalcium silicate is red, dicalcium silicate is blue, tricalcium aluminate is bright green, tetracalcium aluminoferrite is orange, gypsum is pale green, free lime is yellow, and potassium sulfate is white.


photo here


Figure 6. Final processed image for a second cement.


Next: 3-D Reconstruction Up: Main Previous: Introduction