X-ray computer tomography data processing─ an important step towards the qualitative assessment of porous media
O 12.4 in Thermal groundwater utilization
14.04.2016, 11:00-11:15, Audimax A, Geb. 30.95
Abstract. Over the last decade X-ray computer tomography (XCT) has provided a non-destructive means for visualising and quantifying porous material in 3D. This has enabled researchers and engineers to numerically determine the petro-physical, hydro-physical properties directly from XCT images of soil and rock samples. Despite its high spatially resolved data quality, processing the XCT data and extracting reliable/accurate information is a challenging task. Accurately segmented 3D XCT images can further assist in parameterising digital rock physics models (DRP) which simulate transport properties such as permeability tensor and provide insight into physical phenomena such as distribution of multi-component fluids, or Haines jump mechanism which cannot be measured in laboratory. Three different techniques - namely unsupervised, supervised, and ensemble clustering techniques - were applied to segment XCT rock images and estimate porosity in different rock types. The obtained results were compared with laboratory measurements. Based on sensitivity studies the accuracy and speed with which the above mentioned machine learning techniques performed segmentation and classification the XCT images were compared. Further more geometrical pore size distribution, and effective bulk permeability was calculated using pore network modelling to access the reliability of these segmentation/classification schemes.
Export as iCal: