Predictive digital rock physics without segmentation

doi:10.1016/j.cageo.2021.105008

Predictive digital rock physics without segmentation

2022-02-01

5

PRE

AI

DOI：10.1016/j.cageo.2021.105008

原文链接

原文求助

摘要

En 中文

Measuring physical properties from rock samples is necessary in geosciences to calibrate models from geophysical surveys. Digital rock physics is one way to estimate these properties. X-ray computed tomography (CT) images can be used to create 3D numerical models of rocks. Numerical simulations on such models are proxies for tests performed in the lab. Commonly, segmentation is considered an essential digital rock physics processing step, where each voxel in a 3D model is assigned a property of a mineral phase or pore fluid. Any errors in this process are carried forward and affect all estimations that follow (including density, porosity, permeability, and elastic properties). The density and porosity analyses from segmentation are not predictive, as they are typically calibrated to lab tests. We explore a method that does not use segmentation and instead preserves the scaling relationship between the voxel values that are originally recorded in units of CT attenuation. The method is segmentation-less. Phantoms of known density were scanned alongside our sandstone samples and used as calibration points to convert from CT attenuation to density. A porosity model can be created as this property is negatively related to density. We use effective medium theory to assign a bulk and shear modulus to each voxel. We then experiment with several wave velocity models to estimate P and S-wave speed in order to be most similar to laboratory measurements. The ratio of our P-wave estimations to the laboratory measurements is an average of 0.98 with one effective medium theory. The density and porosity estimates do not require any external laboratory calibration. Overall, the method yields densities with an uncertainty of 44 kg/m(3) and porosities with an uncertainty of 1.67%. This method allows rock properties to be estimated quickly and accurately, on large samples, rare samples, and uniquely shaped samples that may not be the right shape for standard laboratory equipment. The technique is also less arbitrary than segmentation based digital rock physics, and is not as invasive or cumbersome as some laboratory techniques.

Keyword:

Computed tomography

Digital rock physics

Numerical modeling

Density

Porosity

Ultrasonic velocity

Predictive digital rock physics without segmentation

摘要

期刊

机构

引用论文