AI Can Solve Challenges in the Steel Industry in a New Research Project
The University of Skövde, Sandvik, and SSAB are involved in a new research project, with financial support from the Knowledge Foundation and support from the Swedish steel producers' association Jernkontoret. The project includes comprehensive production data analyses that are unique for the field. These analyses will hopefully show new cause-and-effect relationships – which can lead to more efficient and sustainable steelmaking processes.
Analyses made in this project will hopefully show new cause-and-effect relationships – which can lead to more efficient and sustainable steelmaking processes. Picture: Sandvik
The researchers at the University of Skövde, Sandvik, and SSAB will analyse large amounts of data by using AI, Big Data and machine learning during the three years of the project. Today, there are already comprehensive measuring processes in place in the steelmaking industry, but the basis for this project is to find out what information can be analysed if you include all data available for a specific manufacturing process.
– Internet-based retail and economic analysis are examples of sectors that have used advanced data analysis for many years. There is a tremendous potential in the manufacturing industry, and by using machine learning, we hope to find correlations that have not yet been discovered. Correlations that can contribute to solving some of the challenges the steelmaking industry is facing, says Gunnar Mathiason, Lecturer in Data Science at the University of Skövde.
Recycled Steel - a Matter for the Future
Sandvik Materials Technology uses a large percentage of recycled steel, and by doing so it is possible to ensure that the steel manufactured has the right composition with the minimal addition of pure metal alloys. Producing steel of optimum quality and at the same time applying a cost effective and sustainable production process is challenging. This is a challenge for the entire steelmaking industry, and not just for Sandvik.
– During the project, we will analyse large amounts of data through machine learning together with researchers from the University of Skövde. Our goal is that the project will give us a better understanding of our recycled steel categories and our residues. When we have a better understanding of those things, we can improve our optimization calculations, and that way we will be able to reduce the proportion of pure alloys and use more recycled products. If we can achieve that, the effects will benefit our finances as well as the environment, says Magnus Josefsson, Head of Raw Material Optimization at Sandvik Materials Technology.
Increased Efficiency and Reduced Emissions
Steelmaking at SSAB takes place at high temperatures. When carbon meets oxygen, carbon dioxide is generated, which contributes to global warming. SSAB's goal is to have a fossil free steelmaking process by 2045. As a step towards reaching this goal, the company wants to work with the University of Skövde to analyse and create a model for the steelmaking process in BOF with the purpose of delivering optimum temperature and composition, while using a minimum of resources. This task is currently the responsibility of the operator and requires extensive experience. Such a model enables the steel to retain its high quality while emissions, the required amount of energy, and residual materials will be reduced.
– We believe that all levels will benefit from this project. By creating a computer model using AI, we can support operators with less experience. We also believe that the model may find important but previously unidentified parameters for additional improvements in the steelmaking process, says Niklas Kojola, Senior Specialist in steelmaking at SSAB.
Contributions to Developing the Analysis of Big Data
The University of Skövde also hopes to develop new knowledge that will contribute to improved algorithms for complex process analyses.
– Through our efforts in the project Swedish Metal, we will be able to develop the field of data analysis in general and specifically machine learning, says Gunnar Mathiasson.