Often, sustainability and economic profitability are seen as incompatible goals. In reality, they can go hand in hand; manufacturing industries are a clear case in point. In a recent webinar we hosted with Swedish steel manufacturer Uddeholm, experts from both companies discussed how artificial intelligence (AI) can address sustainability challenges in manufacturing. This article shares excerpts from the discussion with a focus on Uddeholm’s use of predictive analytics to improve steel production.

Can AI contribute to sustainable manufacturing?

The core focus of sustainable manufacturing is to ensure efficient use of resources, minimize waste, reduce energy consumption and prevent equipment faults that can cause dangerous emissions or production delays. Industry 4.0 technologies (wireless connectivity, IoT, sensor technology and automation) play a central role in helping manufacturing realize their sustainability objectives. All these technologies help to generate, gather, track and utilize data that forms the bedrock of AI initiatives supporting manufacturers’ business and sustainability objectives such as:

  • Improved quality and process control
  • Predictive maintenance
  • Agile robots that act with greater precision
  • Production planning and logistics optimization
  • Waste minimization
  • Product design
  • Energy efficiency

We are just at the beginning of understanding how far AI can go to support sustainability. However, it is clear to see that the advancements and possibilities this technology offers are enormous. It’s also clear that using AI to meet sustainability and business objectives offers both economic and resource efficiency gains. Manufacturing leaders like Uddeholm are exploring and testing initiatives to assess how best AI can support their visions.

Seeking out interesting AI applications

Founded in 1668, Uddeholm produces high alloyed tool steel used to shape a variety of products in our everyday lives — from PET bottles to spectacle frames. The steel must be of the highest quality, clean and polishable.

As a forward-looking manufacturer, Uddeholm was keen to identify and evaluate areas to apply AI. One opportunity Uddeholm and CGI experts identified was understanding why cracks in the steel occur.

“At that time, we could only detect cracks in steel in the last stage of production. If cracks were found, you had to put the steel back in the oven and melt it again, something that became both expensive and energy-consuming. Because there are so many different factors that can contribute to crack formation, conventional analysis methods were far too extensive. We concluded that the area was well suited for AI testing,” explains Ola Axelsson, Uddeholm's Engineering Manager.

Over a period of one year, data from two common types of steel and their manufacturing process was collected. Subsequently, a machine learning model was developed that analyzed individual parameters and combinations of parameters from the data that could affect crack formations.

Data is an eye opener

The analysis revealed several interesting insights. ”We found that if a certain alloy was within a certain limit and the time for the forging press exceeded 5000 seconds, we ran an 87% risk of the steel developing cracks," explains Ola.

“This was an eye opener for us. Despite CGI’s data scientist having no experience of steelmaking, he obtained crucial information simply by analyzing our data. Finding these connections through conventional methods would have taken years.”

With information about the root of the problem, defective material could be scrapped earlier in the process, saving time, resource and energy. In addition, steps in the manufacturing process could be modified to avoid cracks in steel. In this way, it was possible to both minimize costs and improve sustainability by using less materials, energy and resources.

Getting organized for success

Ola says this initiative revealed the enormous benefits data offered and the importance of maintaining high-quality data. As a result, Uddeholm is setting up a governance model together with our experts to secure the entire chain of data, from collection, storage and quality assurance to analysis. The steelmaker is also setting up a future-ready platform to manage both stored and streamed data so that employees in production can track and act on the information in real time.

Ola emphasizes the importance of shaping the organization to support AI initiatives. Uddeholm has built a "hub" with AI and IT infrastructure experts as well as specialists from different parts of their business, such as production, logistics, quality, etc., to share information on potential areas to work with. "In this way, we are confident that we will solve real problems that we have in the business. This has proven to work very well," he says.

“Our goal is to become a data-driven business where decisions are made with the support of data as far as possible. Data-driven thinking should permeate the entire organization so that it’s possible to spot problems on the floor. That's when things can really happen.” 
Ola Axelsson, Uddeholm's Engineering Manager

Together with Uddeholm, we are working on several other AI projects for increased sustainability, energy efficiency and reduced maintenance costs, including:

  • Predicting the right time to replace the regenerator package in an oven to avoid unnecessary equipment downtime
  • Finding leakages in the compressed air system in real time