Chapter 2

Using advanced data analytics to get more from drilling data: applying data science and AI methods on core image data

By Parham Adiban, Consultant (Data Science/Geology), SRK Consulting

Efficient data collection is crucial for making informed decisions about resource extraction and management. One of the emerging techniques in this field is the use of AI methods such as image segmentation and convolutional neural networks (CNNs) to automate the process of logging core images. This technology aims to optimise how geologists and mining engineers analyse drill cores, by allowing them to collect quantitative and semi-quantitative data that has not been collected or cannot be collected by logging geologists. Automated logging is faster and more cost effective relative to a logging geologist or a photo relogging campaign.

Traditionally, logging drill core is a labour-intensive process that involves visually inspecting each core and manually recording observations. However, recent core logging studies, including SRK led projects, have demonstrated the potential of using photos to extract valuable information directly from the image data. This approach leverages machine learning techniques to automate various aspects of core logging and augment the work of geologists.

Instance segmentation techniques can be employed to identify every vein along a core and quantify veining intensity within geological domains. Similarly, these techniques can detect breaks and rubble zones, providing insights into zones of deformation. This information is crucial for understanding the structural integrity of the rock and planning safe and efficient mining operations. Instance segmentation can also be used to assess the core damage index, which is important for geotechnical evaluations. By automatically identifying damaged sections of the core, engineers can make more informed decisions about the stability and safety of the mine.

Image classification models, such as ResNet, can be trained to identify different lithologies, alteration and foliation intensities from core images. This automated classification helps in creating more detailed and accurate geological models, which are essential for resource estimation domaining and mine planning.

Despite the promising advances, there are challenges associated with the variability in data collection. Factors such as core size, core box material, photography tools, lighting conditions and image distortion can all affect the quality of the data. Additionally, historic data collected using older methods and equipment may not be directly compatible with automated logging techniques.

To address these issues, there is a growing effort within the industry to standardise data collection techniques. Using specific photography equipment to minimise distortion and incorporating colour scales to correct for lighting variations are steps in the right direction. Moreover, tools are being developed to make corrections to historic data, ensuring that it can be used effectively with modern automated logging systems.

In conclusion, the integration of image segmentation and CNNs into geological core logging is positively affecting the mining industry. While challenges remain, the collective effort to standardise data collection and improve data quality is paving the way for more efficient and accurate geological analysis. As these technologies continue to evolve, they hold the promise of further reducing costs and enhancing the overall efficiency of mining operations.