Skip to content

Using ML to Make Green Steel

By: Tim Eschert Tim eschert
Bryan goff b Szx H ATSWQ unsplash

Tim Eschert from Fero Labs talks with MarketSteel about the green steel trend, challenges in steel recycling, and how machine learning can reduce production emissions and costs.

How important is green steel for the industry and where do the challenges lie?

Quite frankly, there is no alternative to green steel--especially if you are serious about decarbonizing the industry. Steel that fails to comply with regulations on CO² emissions, sooner or later, will no longer have a market.

As a result, entire business models are at stake. The steel industry now faces the challenge of redesigning an extremely energy-intensive production process to meet society's demands for sustainability. This also applies to other energy-intensive industries, such as the cement industry.

The necessary transformation will not happen overnight. Rising energy prices and the discussion about a possible withdrawal of gas and coal from Russia have exacerbated the situation, especially in Germany. Compounding these problems are the shortage of skilled workers and the backlog of innovation and investment.

How does ML make steel more climate-friendly?

The road to genuine green steel is long and complex. A steel mill is not a small production plant. Artificial intelligence and machine learning can help reduce complexity by fully exploiting the possibilities of automated data processing in production, where there is still a great deal of untapped potential.

One thing is important to remember: ML is not magic. This is a statistical discipline that can be used in a very predictable way to reduce emissions and cut costs while maintaining quality. The way it works is quite unromantic: The software calculates, very quickly, how to change parameters in the production process to keep the steel within specifications.

Why is green steel interesting for Fero Labs?

We love steel. Steel is a high-tech product with a multitude of highly complex parameters. Recycled steel gets even more complex, as different alloys must be added to maintain quality.

This is where we come in. Fero’s white-box ML software adapts to real-time fluctuations in raw material composition, significantly reducing the likelihood of scrap in the production process. This saves both raw materials and energy. Being able to demonstrably contribute to making steel more climate-friendly motivates us immensely.

What specific projects are you working on?

Together with our partner Gerdau, we are working on influencing the variance in the production process in real time, such as by optimizing the use of raw materials. The project proves: Those who have their process parameters under control significantly reduce waste and have a clear competitive advantage. In Gerdau's case, we achieve savings of around 9% per year for an average production volume.

Where do you see the green steel trend headed in the next few years?

There will be much more precise differentiation when talking about "green steel" or "steel recycling". In addition to upgrading to the most efficient production facilities possible, the steel industry must also continue to exploit the untapped potential of process data in order to reduce the carbon footprint and meet regulatory requirements. ML can--and will--be a key success factor here.

Originally published in MarketSteel.