– When the world is being electrified, there is an increase in demand for copper. Inside a petrol car, there is around 20 kilograms of copper. Inside an electric car, there is 70 kilograms. To meet that increased demand in a safe way we need automation, says David Degerfeldt, Program Manager Artificial Intelligence in Mining at Boliden.
The first step is remote control. This allows operators on the ground to control the mining equipment deep down under the earth's surface. This is made possible in part by a 5G network inside Boliden's mines.
– The second step is using AI and other technologies to make the machinery fully autonomous. We hope this will result in less wear on the machines, and at the same time also increase the production to the level that's needed.
In a presentation given at a webinar arranged by Swedish Mining Innovation, David Degerfeldt recently referred to Global Mining Guidelines Group's (GMG) Maturity Model. The company is currently at the second level out of five, at the one GMG calls the Foundation.
– We have understood AI's potential and done our first case studies, about both predictive maintenance and predicting disturbances, says David Degerfeldt and adds:
– The later case was the more successful one. We could predict disturbances up to two hours before they happened by using AI. Implemented in production, that would be very beneficial because it would give the operators ample time to take precautionary actions and hopefully avoid the problem.
Going forward, David Degerfeldt's goal is to build a company-wide acceptance for AI and how it can be applied within Boliden. The over-arching vision is a data-driven Boliden.
– We have a lot of work to do to build that acceptance and understanding. Making data available and using it for our decision making is a good starting point, I think. And when the data volumes grow, the need for AI will come.
As one example, David Degerfeldt mentions a project based on reinforcement learning on one of the primary mills in Aitik.
– Training the model on our simulation of the mill revealed that it would take 200 years before the models would perform better than the controller we use today – if the model was trained on real-world data from the mills. That highlights why simulations are so important: You want to use the technology where it makes an acutal difference, and by training the models on simulations you can get there faster.
In the end, David Degerfeldt is certain that AI will have a great impact on how Boliden operates, and by that also on how the company is organized. When the toolset changes, the competences needed change as well.
– For a long time, our heroes have been down in our mines. When our machinery is controlled remotely, and eventually fully automated, we will have new heroes in new positions.
Why has Boliden decided to join AI Sweden as a partner?
– We hope that AI Sweden can help the whole industry to move faster and further in this journey. It is to complicated for any individual company to make the transformation on its own. AI Sweden has an important role in coordinating and connecting all great intiatives both withing the private sector and academia.