The ALCHIMIA Project: Utilization of Explainable AI for Smart, Sustainable Steel Production
By Sandra Rothe, VDEh-Betriebsforschungsinstitut (BFI)
In the rapidly evolving world of steel manufacturing, efficiency and precision are essential to staying competitive. As industries adopt cutting-edge technologies like artificial intelligence (AI), the need for trust and transparency in AI-driven decisions has become more important than ever. This is where Explainable AI (XAI) comes into play, and at the Alchimia Project, we are using it to optimize operations and drive the future of smart, sustainable steel production.
What is Explainable AI (XAI)?
Explainable AI is a branch of AI that focuses on making the decision-making processes of AI systems understandable to humans. While traditional AI models often act as “black boxes,” providing answers or recommendations without clear explanations, XAI shines a light on the why behind these decisions. It helps us see exactly how and why AI reached a particular conclusion, making the technology more transparent, trustworthy, and actionable.
In steel manufacturing, where complex processes involve countless variables, transparency is critical. Companies not only need to benefit from AI’s predictions but also understand the reasoning behind them. This is where the Alchimia Project bridges the gap, making AI decisions more accessible and understandable.
How the Alchimia Project Applies XAI in Steel Manufacturing
Using scrap in steelmaking is a key component in supporting the circular economy by recycling and reusing materials. It conserves energy and resources, offering a more sustainable alternative to using virgin iron materials. However, this comes with challenges, such as the fluctuation in composition, availability, and pricing of different scrap types over time and between suppliers. Additionally, inconsistent metallic yield makes it difficult to ensure consistent quality, and residual elements like copper (Cu), which cannot be removed from liquid steel, further complicate matters.
One of the main challenges is predicting steel quality, which involves determining the content of various elements in the produced steel. This requires a tool that not only predicts steel quality but also helps to understand the individual element concentration from each scrap type. While traditional scrap characterization tools use statistical methods, AI-based tools offer more accurate predictions of steel quality. However, predicting scrap composition with AI is more difficult because there is often insufficient training data to build reliable models. This is where XAI becomes critical—it helps us better understand the AI’s “black box” and clarify how it predicts steel quality, enabling us to draw clearer conclusions about scrap composition.
Building Trust with XAI: The Path to Sustainable Steel Production
Sustainability is a growing priority across all industries, and the steel sector is no exception. By incorporating XAI into production processes, the Alchimia Project is paving the way for smarter, greener manufacturing. With every AI recommendation backed by transparent reasoning, steel manufacturers can confidently make decisions that improve efficiency while reducing waste and energy consumption.
As the industry moves toward a future where AI plays a crucial role in every aspect of production, explainability will be essential for maintaining human oversight and trust. XAI ensures that AI is not just a powerful tool, but also a responsible and ethical one, empowering manufacturers to make sustainable choices that align with both environmental and economic goals.