How AI supports green steel production

Source: https://www.voestalpine.com/blog/de/innovation/sehr-gescheit-kuenstliche-intelligenz-bei-voestalpine/

Artificial intelligence (AI) is gaining ground in the dynamic landscape of industrial manufacturing. A tool not only to increase efficiency but also to reduce environmental impact. In this context, AI is becoming increasingly important in the production of green steel.
Here we look at the key role AI is playing in the transformation of steel production and how intelligent technologies are making the steel industry not only more efficient, but also more sustainable.

© Source: https://www.mpie.de/4894435/ml-for-corrosion-science

a) Schematic representation of the neural learning model b) Schematic representation of data processing using natural language processing. LSTM: long short-term memory. © Image taken from Science Advances, DOI: 10.1126/sciadv.adg7992

AI as a key player in highly specialised production

AI-based applications are spreading rapidly. Experts are facing a veritable flood of new trends and tools that are fundamentally changing the rules of the game in manufacturing. The use of AI in industry goes far beyond simple automation. It acts as a brain that processes information, recognises patterns, solves problems and makes decisions in a matter of seconds. The industrial landscape is dotted with examples that illustrate the use of AI in quality inspection, predictive maintenance planning, and monitoring and control of production processes.

Foundations for high-performance AI solutions

The performance of AI-based solutions is closely linked to the data input. AI models need to be fed with sufficient amounts of high quality data to provide the appropriate output. In manufacturing, there is usually a wealth of accurate sensor data that can be used to train AI models.

AI is also used in materials science to predict the corrosion behaviour of materials, for example, to find optimal alloy compositions. However, the predictive power of AI models is limited, as not all relevant data can be taken into account. Typically, such AI models process numerical data. The production and processing of an alloy, on the other hand, is documented in text form. Scientists at the Max Planck Institute for Iron Research (MPIE) in Düsseldorf have now succeeded in developing a machine learning model that can process both numerical data and natural language fully automatically. The model can be used to predict corrosive failure with 15% greater accuracy. The model will now be extended to include microscopy images so that all relevant sources of information, text, figures and images can be incorporated into the AI model to further enhance its predictive power.

© Source: https://www.primetals.com/de/presse-medien/news/neuartige-ki-basierte-computer-vision-loesung-von-primetals-technologies-reduziert-stillstandszeiten-bei-thyssenkrupp

The digital assistant recognises belt cross breaks with AI and immediately reports errors to the operating personnel.

Where efficiency meets sustainability

Of particular interest to steel production is AI's ability to compare parameters from current and historical production processes and calculate the optimum present condition in fractions of a second. At the forefront of this development is ArcelorMittal Hamburg, which is using AI to optimise the trimming process in wire rod production.

Another key feature of AI-based solutions is their ability to detect process phenomena at an early stage and make recommendations. This enables proactive identification of faults, thereby increasing overall efficiency. Operators can act quickly to correct deviations and restore normal operations. For example, Primetals Technologies implemented a digital assistant at thyssenkrupp Hohenlimburg to detect strip cross-cracks in rolled intermediate products. This is the first computer vision-based solution developed by Primetals Technologies specifically for this purpose. A camera on the roughing line provides a live video stream to the digital assistant developed by Primetals Technologies. Using a machine learning model, the assistant processes the live images and checks them for cracks. If a crack is detected, the staff receive a corresponding message immediately. The digital assistant also transmits the information to the customer's intelligent automation system for further analysis.

Advanced Raw Material Tracker (Source: SMS Group, https://www.sms-group.com/services/bfxpert)

SMS group, in turn, optimises blast furnace operations with its process control software "BFXpert". The Advanced Raw Material Tracker, a three-part upgrade of BFXpert, uses real-time control models for raw material tracking: The transport model monitors bulk material during the process and enables more efficient use of conveyor belts. The hopper model provides a real-time calculation of material distribution based on material properties as the material hopper is filled and emptied. The load calculation model provides information on the radial distribution of each material component in the blast furnace to optimise belt utilisation. These models have already been successfully integrated into production operations at two blast furnaces at Dillinger Hüttenwerke, enabling early detection of faults.

Software CO2 AI, Boston Consulting Group, Screenshot from https://co2ai.com/

voestalpine Grobblech GmbH also relies on AI. It has commissioned nokra Optische Prüftechnik und Automation GmbH to introduce an optical measuring system for the final inspection of roll-bonded clad plates. The high-resolution measuring system enables precise thickness measurement and has led to a significantly higher throughput. It not only records the external geometry, flatness and surface quality, but also measures the thickness of the plates over their entire length and width using light section sensors. The accuracy of the thickness measurement is +/- 100 μm, while the flatness meets the requirements of DIN EN 10029. Whereas it used to take one or two shifts to manually measure a sheet, the new system provides information from more than a million measuring points within ten minutes. In addition to improved quality assurance, the precise data can also be used to optimise processes, as the digitalisation of the process enables all results to be automatically stored and evaluated. This is another example of how artificial intelligence can significantly improve the efficiency and quality of steel production.

AI-supported CO2 balancing for steel products

The advantages of green steel are not only of an ecological nature, but also offer great opportunities for companies. Customers will increasingly pay attention to how climate-friendly a product is and what contribution it makes to reducing greenhouse gas emissions. Companies that invest in green steel at an early stage can therefore realise greater market shares and higher profits. The Boston Consulting Group (BCG) has calculated that demand in Europe will exceed supply by up to 20 million tonnes by 2030.

So far, however, there is no internationally accepted definition with standardised threshold values as to when a steel or raw material can be called "green". One important aspect is the measurement of the CO2e footprint along the entire value chain. The so-called Product Carbon Footprint (PCF) includes all greenhouse gases that are generated along the entire value chain of a product - from raw material extraction to production and distribution. The Scope 3 emissions of each individual process step are also included.

Although the Product Carbon Footprint (PCF) is currently used only by a few companies, it will play an important role in the future. However, calculating the PCF is complex and requires compliance with globally recognised standards. One way that steel manufacturers and steel suppliers can accurately calculate and track the carbon footprint of their products is through the Boston Consulting Group's CO2 AI software. CO2 AI has the ability to cover the entire life cycle of a product and utilise real-time data. This enables the software to provide a complete balance sheet from raw material extraction through production to delivery, thereby increasing transparency towards customers and the public. This example shows how AI supports solution finding and helps to determine and quantify the CO2e footprint. Metal distributor Klöckner & Co. has been relying on such an AI solution since January of this year, which was developed in collaboration with BCG and tailored to the company's needs: On request, the certified Nexigen PCF algorithm provides Klöckner customers with a Product Carbon Footprint Declaration for each order, in which the greenhouse gas emissions generated are specified to the nearest kilogramme.

The PCF can become an important value driver as customers are increasingly willing to pay more for green products. Companies that are early adopters of this parameter can benefit from increased market share and shareholder value, as PCF is expected to become a key driver of purchasing decisions in both the B2B and consumer markets. The transition to green steel requires an adaptation of the value chain, and AI solutions play a crucial role in this by not only increasing efficiency but also supporting environmental sustainability.

The successful integration of AI into steel production will therefore pave the way not only for green steel, but also for a sustainable and value-driven future for the industry.