Learning from Machine Learning: Key Takeaways for Machine Learning Applications to Rock Mass Characterization and Classification

Abstract

ABSTRACT: As the scale of mining projects increase, rock engineers now face the task of analyzing vast amounts of data collected through traditional methods (boreholes) and newer methods (remote sensing). Machine learning (ML) has emerged as a tool that can assist rock engineers with this, allowing for automating and improving rock mass characterization and classification methods. ML has existed for decades, but its application to rock mass characterization and classification is more recent and has significantly increased over the past few years. The adoption of a different technology comes with a learning curve, where the beginning is marked with inflated expectations and implementations that are not always practical (such as using complex ML models when working with limited geotechnical data), before eventually reaching a stage where its applications are useful and widely adopted among rock engineers. This talk will take a critical look at the past and current state of ML applications to rock mass characterization and classification. It will highlight successful use cases and draw key lessons that can help guide the future development of ML applications in rock engineering.