Automating Rock Quality Measurements Using Core Images and Machine Learning
Federico Arboleda (Co-founder of Imago)
Previously unexplored workflow using geoscientific imagery is starting to be considered essential because they can speed up our efforts to define geological and physical properties without the overhead of accessing the material physically. This situation has produced a paradigm change on where we collect drilling data and how we view its value.
A timely and accurate geotechnical model is essential for mine planning and design. Drilled core samples are a valuable source of information for rock mass classification for the purpose of predicting rock mass behaviour. Machine learning working together with consistent core photos is making RQD measurements faster and more reliable.
Typically drilling core measurements involve a labour-intensive visible inspection of the core samples at the mine or exploration site – incurring significant cost. The process is time consuming and consistency is dependent on training and experience of the operators.
Advances in low cost cameras and image-based machine learning have led to recent experiments in automating the measurement of RQD from drilling core samples with the goals of:
* Faster, lower cost geotechnical measurements from core samples.
* Improved repeatability and accuracy.
This paper presents a new technology framework and process for automated RQD calculations from core images based on machine learning and reviews the results.
