Machine Learning: A New Approach to Familiar Problems
As a recent series of technical posts made clear, resource estimation is crucial for defining life of mine and mine planning, and is the cornerstone of the mining business value chain: a tiny error in the estimation can lead to the loss of millions of dollars. So how can we make resource estimation — along with other critical mine processes — more reliable?
One potential answer lies in machine learning.
What machine learning is
Machine learning is a branch of artificial intelligence (AI) that uses data and algorithms to imitate how we as humans learn: through experience and adaptation until gradually we get better and better at whatever task or skill it is we want to master.
If you own a Netflix account, you’ve experienced machine learning in action. It’s what the company uses to make personalised recommendations to its millions of customers. Self-driving cars are also based on machine learning, and it has made image recognition, text translation, and virtual personal assistants possible.
Resource estimation and machine learning
Machine learning is also gradually making its way into resource modelling and estimation. Several recent pilot projects have used the technique to locate and estimate mineral deposits by training algorithms with large datasets to identify patterns not detectable by humans alone — a critical advantage at a time when it’s becoming harder and harder to find new near-surface mineral deposits anywhere in the world.
In 2020, for example, a team of researchers from the Australian Institute for Machine Learning at the University of Adelaide combined government magnetic and gravity data with AI analysis and machine learning to help geologists visualise and map the features of rocks in undiscovered mineral deposits deep underground. Elsewhere, a group of resource geologists developed machine learning models with historical geophysical data to establish orebody properties for non-core drilling holes: the project generated critical properties and lithology predictions with confidence as high as 95 percent.
Mining professionals are also beginning to use machine learning to improve accuracy and efficiency in other mining processes. UK-based Ionic Engineering, for one, has employed it to enhance image recognition used to identify copper grades and increase product quality, while Canadian mining consulting firm Shyft Inc. developed machine learning algorithms to forecast energy peaks as part of a project to enhance automatic ventilation systems control and reduce energy costs in underground mines.
However, it is important to remember that these applications are relatively recent and that innovation is always a challenge.
For example
Implementing new, sophisticated technology takes a leap of faith, and the mining industry as a whole is traditionally risk-averse. Of course, we have good reason to be: the stakes are incredibly high and no one can afford to waste money. Sticking with the tried-and-true is safer than jumping on board even the most promising innovation.
Plus, even if a company is prepared to adopt sophisticated technology, implementation may be difficult, either because its mining projects are spread over a huge area, with thousands of kilometers of roads between mining sites and ports, or based in remote locations without the required infrastructure, or because its own mining professionals are not yet up-to-date with the latest advances.
However, these are not insurmountable problems, and it is both possible and desirable for mining companies to retool and retrain their organisations to use machine learning and other AI solutions to take advantage of large amounts of (often-unexploited) data — because this data may well hold the information they need to improve mine optimisation and increase profitability.
This series
The resource estimation series posted by Michael Mattera made the point that advances in data acquisition technologies mean that a whole new world of data — from downhole geophysics, multi- and hyper-spectral core scanning, and more routine collection of geometallurgical parameters — is now available to inform and enhance resource modelling and estimation.
Because the acquisition of big data is also quickly moving ahead for other mining practices as well, the next step for the mining industry as a whole has to be learning how to turn all that new, raw data into useful intelligence. Machine learning, in particular, holds great promise for improving the reliability and efficiency of mining processes.
The following series of three articles discusses how mining companies can begin to integrate machine learning into their work along with detailed tips for how to use it to solve real mining problems.
Related posts:
