Artificial Intelligence is the Next Frontier of Improving Underground Mining Operations
What is Artificial Intelligence to Mining?
Artificial intelligence IS the next frontier of improving underground mining operations. But, we’re not talking about turning your mine into a hive of drones. This isn’t about upending your operations. Machine learning (ML) algorithms pull out patterns in data collected around your mine and fleet to predict problems before they occur. This vastly improves the efficiency of your mine.
“What machine learning algorithms do is pull out patterns in data collected… to help predict problems before they occur…”
At Agnico Eagle’s Goldex Mine near Val d’Or, Que., they’re saving on maintenance costs for the engines of their haulage trucks. This is thanks to the predictive analysis made using their historical data collected by Newtrax and applied to ML algorithms.
The maintenance team at Goldex discovered that their truck engines were sometimes failing, a lot earlier than expected. The team had suspicions on what was causing the problem and were alerted by the Newtrax Mobile Equipment Telemetry (MET) system. This provided them with a good indication. However, by using ML, we were able to sift through the data collected from the truck’s engine sensors far more quickly. We were able to pick out the patterns of what was happening with confidence resulting in faster time to action.
“[Newtrax] developed a system that enables us to predict issues at least two weeks in advance, even before the alarms, so we can intervene before we incur problems that break our engines,” Agnico’s Daniel Pinard says.
Rather than spending $100,000 to replace an engine, they only had to spend about a quarter of that to make the necessary repairs. And at the same time were able to leave the truck in production, where it moves about 500 tonnes of material a shift.
This is what AI is all about. Finding small, manageable solutions that save costs and optimize processes.. Newtrax combined its in-house advanced ML expertise with the client’s data to achieve quick wins.
“It’s not just engine failures that machine learning algorithms can predict in advance, it can see failure in smaller components, too.”
It’s not just engine failures that machine learning algorithms can predict in advance! ML can see failure patterns in smaller components, too. Take batteries, as an example. Typically during normal usage, a battery puts out a precise voltage that holds steady for its entire life. But as a battery ages, and approaches failure, subtle variations start to appear.
Newtrax engineer Louis-Pierre Campeau saw the algorithm was picking up a change in the output. When he looked at it, and said, “Well, it’s 26.7 Volts, and it was 26.7 before so it’s okay.” But the algorithm was saying there was a problem.
“Then looking deeper, the average was 26.752, and afterward it was 26.716,” Campeau says. “It’s a very small change you just couldn’t pick up by looking at it.”
“This is the REAL power of machine learning. A single person isn’t able to look at all the data… but an algorithm can…”
This is what he says is the real power of machine learning. A single person isn’t able to look at all the data for the different measurements given off by a fleet of vehicles. But an algorithm can, and it’s able to draw connections of things a person isn’t likely to even see as related.
“Sometimes there are too many factors put together that you can’t really look at as a whole,” Campeau says. “If one measure goes up while another goes down, at this average and the temperature is going down. There’s a certain limit of information you can process out simultaneously.”
“Newtrax has a large collection of anonymized data…”
What Newtrax brings is a focus on underground mining. You won’t be a client with a company that’s adapting their algorithms for the train yard to fit the underground environment. Newtrax has a large collection of anonymized data from our client base of underground mining companies. We’re able to draw on right from the start, Campeau says
“Each mine is part of the large pool of clients we have, and can use all of these insights we get from our previous experience to apply it right away,” he says. “Rather than having to figure out everything from scratch.”
“You may not have seen your LHD’s in every possible situation, but there’s a good chance we have, and can pull patterns from it right when we start.”
We’re able to draw on a large database collected from a number of different mines, that gives us a huge wealth of real-world data to train our algorithms. You may not have seen your LHD’s in every possible situation, but there’s a good chance we have, and can pull patterns from it right when we start.
For more information on Newtrax successes at Goldex, you can read this case study: