Predictive digital rock physics

Geophysical surveys are the most effective way to model the surface and subsurface. Recorded signals can only be converted to properties of interest, such as rock type or saturating fluid, if we understand the relationship between rocks and their physical properties. Digital rock physics (DRP) consists of creating digital models of rocks in order to numerically model their properties. Computed Tomography (CT) scans are a common starting place for DRP models. Each model consists of voxels (3D pixels), with values in units of CT number, which are an approximation of X-ray attenuation in that location. The conventional method of converting CT-numbers into rock properties is known as segmentation. Users assign physical properties to each voxel based off the visually identified material that the voxel represents. Segmentation introduces arbitrariness causing certain rock properties to be drastically incorrect. Measuring some sample properties in the laboratory for calibration is a common mitigation. Here, we introduce a new segmentation-less DRP method and present case studies suggesting that we can considerably improve the prediction of rock physical properties and match physical laboratory results. Chapter one demonstrates that by scanning objects of known density alongside the rock sample, we can calibrate the rock model. We refer to the method as “predictive”, as sample laboratory measurements are not required. We explain the steps to create density, porosity, and wave velocity models, and derive a method to estimate uncertainty. In chapter two, we acquired CT scans with varying conditions to test the sensitivity of the method. We provide advice and reasoning on how to set up an effective scan. In chapter three, we estimate rock properties from millimetric sized sandstone fragments too small to test with conventional equipment. We also estimate properties for meteorite samples too rare to invasively test. This shows the method is more than academic, and can meaningfully contribute to scientific debate and industry decisions. We have shown that CT, a relatively new technology not widely adopted for this purpose, can be used to accurately solve rock physics problems when we combine knowledge from material science. This more accurate methodology could make this powerful technology more mainstream for rock physicists