Open Collections

Predicting the properties of rock using data-driven methods

University of British Columbia

The concept of digital rock physics (DRP) is widely used in petroleum engineering and science. It is a digital interpretation of a real rock specimen which is used to visualize and model rock properties over multiple scales. It is an excellent cross-over concept for mining. In the oil and gas (O&G) industry, making the most of every centimetre of the core, or gram of cutting sample is imperative when it comes to minimizing the risks and improving the outcomes of an exploration program. Similarly, to O&G applications, DRP can be translated directly into mineral exploration and mining with a significant effect, especially in the field of geometalurgy. By creating a 3D version of a core sample based on real rock properties and not using numerical or probabilistic models, it is now possible to perform a multitude of virtual experiments and observations without damaging the original sample. The most commonly used rock analysis includes high-resolution micro-computed tomography (micro CT), scanning electron microscopy (SEM), and focused ion beam (FIB) imaging, that enables a 3D analysis of the rock's structural and mineralogical properties at higher resolutions compared to that which is possible with light microscopy. However, these DRP methods are relatively slow in data collection and have other problems that limit the use of the technology to our advantage. This thesis presents a robust method for rapid, large-scale acquisition of data from digital models of rock specimens, combined with an automated data segmentation using machine learning, which dramatically increases the speed of digital rock analysis. The feasibility of the approach is demonstrated through the 3D analysis of both homogenous and highly heterogeneous rock samples by achieving a significant improvement in speed of analysis as compared to manual approaches for data segmentation and digital rock analysis.

Text

2021-09-30

Attribution-NonCommercial-NoDerivatives 4.0 International

http://hdl.handle.net/2429/76079

Mining Engineering

University of British Columbia

UBCV

http://creativecommons.org/licenses/by-nc-nd/4.0/