ARMA UToronto  Student Chapter Online Talk:  Dr. Sebastian Goodfellow

Actions Panel

ARMA UToronto Student Chapter Online Talk: Dr. Sebastian Goodfellow

Doing More with the (ATV) Data We Have: Understanding Stress in Deep Mines

When and where

Date and time

Location

Online

About this event

Short Bio

Dr. Sebastian Goodfellow recently joined the Department of Civil and Mineral Engineering at the University of Toronto as an Assistant Professor. Before joining the University of Toronto, Dr. Goodfellow worked for four years at a mining technology startup, KORE Geosystems, and led the development of a machine learning product designed to provide geological and geotechnical core logging assistance to geologists and engineers. The technology won several awards (2019 Victorian Premier's Design Award, 2019 Good Design Award, 2017 Disrupt Mining Winner) and has resulted in government funding and capital investment. Dr. Goodfellow's focus is on the application of new technologies to conventional data sources to tackle challenges in the mining industry and studying the interactions between humans and these technologies in high-stakes mining environments to support successful integrations.

Abstract

Knowledge of local stress conditions is of critical importance for mine design and mine management. However, stress is a challenging quantity to measure because it can be highly variable across a region of interest and the available methods can be time-consuming, expensive, and have high rates of failure. As a result, measurements are sparse and mines may therefore be designed and operated with an uncertain understanding of the stress state.

Over the past decade, new technologies such as Acoustic Televiewer (ATV) for geotechnical and geological logging of boreholes, the Cloud, and Machine Learning (ML) for data analysis have emerged, and the adoption of these has now reached a tipping point in the mining industry. We have observed an 8-fold increase in ATV data collection by the Canadian mining industry since 2014, an increase that has largely been driven by improvements in the technology. As a result, the mining industry is now sitting on big ATV datasets that were gathered for borehole logging, but have potential value that far exceeds this. The opportunity we are exploring is to apply ML techniques to ATV data in order to obtain novel and improved assessments of the local stress state.

As shallow deposits become exhausted, mining targets are extending to greater depth, and are now approaching 2 – 3 km in Sudbury, Ontario. In these conditions, engineering design decisions made without an accurate and complete understanding of stress conditions expose underground workers to the risk of death and serious injury following structural failure and may threaten the overall financial feasibility of a mine. By developing a methodology and tools for robust real-time in situ stress estimation, we will be able to provide decision-making intelligence to mine managers allowing them to make more informed decisions.