Early in his career, Kevin Urbanski was on the team at Matrikon (now part of Honeywell) that built one of the first systems to gather sensor data from mobile mining equipment. Years later, working at Teck Resources, Urbanski was in the position to use that data to assess equipment health—and hit a major challenge. With hundreds of sensors on each piece of a fleet’s equipment, there was almost too much information, making finding actionable insights “like a needle in a haystack.”
That was the driving force behind Rithmik Solutions, the company Urbanski co-founded with Kris Isfeld and Amanda Truscott in 2018. After leaving Teck and taking a sabbatical to study artificial intelligence (AI), Urbanski realized that AI could help find those needles, and began investigating how to deploy it effectively.
Rithmik’s AI-powered Asset Health Analyzer (AHA) develops an optimal working model of equipment based on sensor data from a site’s entire fleet. The model runs at the same time as the equipment, and identifies when certain sensor values deviate from the model. The model is tuned based on historical site data—so the optimal condition for a Caterpillar 793D haul truck might be different at mine sites in Chile or Yukon due to the different seasonalities and conditions, such as temperatures, topographies, road conditions and operator training.
“AI finds optimization opportunities from the get-go,” Urbanski explained. While building its models, Rithmik has unearthed opportunities to improve maintenance practices and operator driving behaviour. Some insights have been more unusual, like identifying a truck driver who tapped the gas pedal to the beat of Metallica songs.
Rithmik now has a home at the Mila-Quebec AI research institute in Montreal, which it partnered with in January 2023, and works with major mining companies and OEMs in North America, South America and Africa.
Urbanski has seen excitement for the possibilities of AI in mining—an industry known for its cautious approach to technology adoption—and some roadblocks. Technology projects require resources, but with labour shortages, mining teams have been stretched thin. Rithmik can help to mitigate that by handling tasks like taking inventory of a customer’s data or integrating new insights into their workflows. Its AHA can also multiply the time of maintenance personnel, reducing the hours required to troubleshoot and find improvement opportunities.
Unlike the hype of five years ago, when Urbanski said people expected “magic” from AI, experience with large language models (LLMs) such as OpenAI’s ChatGPT have given people a more realistic understanding of AI’s capabilities—but more pessimism than may be warranted.
“The biggest obstacle is understanding what’s possible, being open to using [AI and LLMs] and having trust in them,” he said.
Urbanski thinks LLMs represent an opportunity for miners; they could be deployed for specific tasks, like scheduling mobile mining equipment or planning preventive maintenance, or for much larger tasks.
“Mines [store] so much information—sensor data, work orders, [information] from pit to port,” he said. “Optimizing that entire chain is really hard to do with disparate technology. LLM architectures allow you to feed them a tonne of information and can find relationships and optimizations.”
The company closed a $2 million funding round in March, which it plans to use to grow its customer base and launch new products. To date, Rithmik has built models for trucks and dozers, and plans to develop models this year for electric and hydraulic shovels and a scheduling assistant to recommend when it makes sense “environmentally and economically” to send a mobile asset in for maintenance.
June 03, 2024