Quantitative characterization of coal by means of
microfocal X-ray Computed Microtomography (CMT)
and Color Image Analysis (CIA)

Frederik J Simons, Frédéric Verhelst and Rudy Swennen

Fysico-Chemische Geologie & Department of Civil Engineering
Katholieke Universiteit Leuven (KUL)
Leuven, Belgium

Abstract

Microfocal X-ray computed microtomography (CMT) isa novel technique that produces three-dimensional maps of the distribution of the linear attenuation coefficient inside an object. In contrast to the more conventional medical computerized tomography (CT) systems, a microfocal X-ray source is used. This enables a far better spatial resolution. The linear attenuation coefficient or tomodensity is dependent on the physical density and the mineralogy of the object to be imaged, and on the energy of the radiation used. Earlier work by Verhelst et al. (1996) presented the results of the correlation of the tomodensities obtained from three-dimensional CT scans with two-dimensional data on the composition of a coal sample acquired with color image analysis (CIA), a camera technique. This analysis assumed the linear proportionality of the tomodensity with the real physical bulk density, which is true only for certain energy ranges. In this paper, we use CMT-devices for a similar correlation. For sake of comparison, the same core sample was used. First, new CIA-data on the surface composition along two profiles were sampled with greater detail (100 micron). These data are subjected to a geostatistical analysis to quantify the spatial dependence between the measurements. Second, CMT-tomograms were made, yielding spatial resolutions twice as high as medical CT. A multivariate correlation was carried out, and two improved (geo)statistical methods are suggested. The different energy range of the microfocal X-ray source compared to medical CT, however, produces some bias in the correlation of the tomodensities with the surface percentages of the constituents. We therefore suggest that the linear attenuation coefficient be treated as a separate unit. No attempt was made to translate the linear attenuation coefficient to the physical bulk density of the different constituents of coal (e.g. macerals).

Figures

1. Figure 01 Energy dependence of the linear attenuation coefficient for hexagonal carbon
   
2. Figure 02 The microfocal X-ray tube
   
3. Figure 03 Geometrical unsharpness
   
4. Figure 04 Reduction of intrinsic unsharpness
   
5. Figure 05 Schematic representation of Color Image Analysis (CIA)
   
6. Figure 06 Photograph of the surface of the coal sample
   
7. Figure 07 Computerized Micro Tomography tomodensities of the coal sample
   
8. Figure 08 Cumulative normalized surface percentages of the five constituents (vitrinite, liptinite, inertinite, pyrite and binder) sampled along trace 1
   
9. Figure 09 Cumulative normalized surface percentages of the five constituents (vitrinite, liptinite, inertinite, pyrite and binder) sampled along trace 2
   
10. Figure 10 Filtered normalized surface percentages of trace 1
    
11. Figure 11 Filtered normalized surface percentages of trace 2
    
12. Figure 12 Results of the multiple regression for trace 1: predicted versus observed tomodensity
    
13. Figure 13 Results of the multiple regression for trace 2: predicted versus observed tomodensity
    
14. Figure 14 Predicted versus observed tomodensity after breakpoint regression
    
15. Figure 15 Experimental semivariogram for the virtinite surface percentage of trace 1
    
16. Figure 16 Experimental semivariogram for the virtinite surface percentage of trace 2
    

Frederik Simons