Numerical computations from X-ray micro-computed tomography (MCT) images yield dynamic flow, petrophysical, and elastic properties which are conventionally obtained from subsurface logs or laboratory measurements on core plugs. These Digital Rock Property (DRP) determinations in most cases have significant advantages in accelerated turnaround time and lower cost, and in the non-destructive nature of the process. Despite these advantages, several limitations prevent this otherwise promising new field from replacing conventional laboratory measurements. Two primary ones are in the areas of gray scale image segmentation, and extrapolation of properties computed at small sample sizes to larger sizes at which laboratory measurements are made, These limitations are particularly severe for carbonate reservoirs, where pore types and pore networks have greater heterogeneity and complexity than exist in clastic rocks.
This paper focuses on digital texture analysis and classification of MCT images not routinely exploited in conventional DRP applications. We demonstrate that such analysis provides valuable information that improves the quality of physical property computations made from MCT images. In particular, we investigated texture-based image processing techniques that address the known limitations of DRP technology, using multi-resolution MCT images of core plugs and micro-plugs from an early Cretaceous-age supergiant carbonate reservoir in the Middle East. These plugs represent a variety of carbonate facies: grainstones, packstones, and wackestones, with porosities in excess of 20%. We adopted an approach to gray scale segmentation that has a physical basis in separating pores from grains. Our segmentation process yielded computed porosities that are in reasonable agreement with laboratory measured values. To extrapolate rock properties measured at small sample scales to those at larger scales e.g. at which laboratory measurements are typically made, we use textural classification, with textural classes calibrated to measured or calculated rock properties. Textural classification has the added benefit of providing an objective basis for rock typing, in contrast to the subjective industry practice involving descriptions of thin sections. Rock typing is key to the extrapolation of core-derived rock properties in reservoir simulation models, and digital textural classification provides the capability for sensitivity studies during extrapolation.
Several computational rock physics models used in the petroleum industry make simplifying assumptions regarding the pore systems in reservoir rocks. For example, the Gassmann equation used in elastic property modeling assumes homogenous grain and pore distribution, with connectivity of all pores. To better understand the complexity of the pore framework and network in our carbonate samples relative to such assumptions, we extracted from MCT images pore attributes that influence flow, petrophysical, and elastic properties. These include pore shape, size, orientation, throat size distribution, connectivity, tortuosity, and relative volumes of connected and isolated pores. These pore attributes enable the use of effective medium rock physics models that compute elastic moduli. We tested the Xu-Payne Differential Effective Medium Model using pore attributes derived from MCT images of our carbonate samples, and found agreement with the results of more sophisticated Finite Element Modeling, and with results from other theoretical modeling studies of carbonate rocks.
