Multiscale Digital Rock Modelling for Reservoir Simulation

Abstract

We present analysis of sandstone micro-CT scans obtained with different resolution. We show that use of fine resolution (about 1 micrometer per voxel) do not provide valuable information about the core structure and for the pore surface analysis and makes the sample nonrepresentative even for porosity estimation. Scans with resolution of 3–5 μm per voxel allow to get statistically reliable estimates of the reciprocal pore-to-core distribution, topological properties of the pore space and transport properties of the sample. Using this information one can reconstruct the macroscopic model of poroelastic media.

Two‐step morphology‐based denoising and non‐local means smoothing improves micro‐computed tomography digital rock images

“…In other words, most images will have similar patches in the image having similar noise characteristics across the image, which could be used for denoising these similar patches by averaging. Recently, NLM filtering has established itself as a standard pre‐processing step (Andrä et al., 2013; Verri et al., 2017; Sell et al., 2016; Sun et al., 2017; Dong et al., 2020; Reshetova et al., 2020) in the digital rock workflow.…”

“…Filters like anisotropic diffusion (Sheppard et al., 2004) and bilateral filter (Verri et al., 2017) have been a default choice in the workflow in the past as they preserved significant features while denoising. From recent literature on digital rock (Saxena et al., 2019; Andrä et al., 2013; Verri et al., 2017; Sell et al., 2016; Sun et al., 2017; Dong et al., 2020; Reshetova et al., 2020), we observe a switch to non‐local means (NLM) filtering (Buades et al., 2005) as it has proven to be more adaptive and efficient in preserving edges while denoising (S. Berg et al., 2018). Hence, we have considered NLM filtering as the base filter technique to compare our proposed denoising pipeline.…”