1 January 2021

Data enhanced physical models to reduce material use in steel production

A significant part of Dutch (and worldwide) energy consumption and carbon emission is related to the production of primary metals, especially steel. In order to become more sustainable, either less energy must be used to produce the material or less material must be used. Materials reduction can be achieved by recycling, by increasing the materials life, or by reducing the amount of material required through improved properties and design. A quest for a radically improved materials modeling approach.

A game changer in materials modeling

Steel produced from ore has excellent quality and can be produced within strict tolerances, which are prerequisites for subsequent steps in today’s manufacturing processes. For materials reduction in producing steel for high-end applications, we need to improve the structural properties, recycling rates, and durability of materials by tens of percents.

Using less energy

Significant energy savings can be achieved by producing metals from scrap (recycling) instead of ore. However, the higher the content of scrap, the more difficult it is to maintain the required quality standards and tolerances. This hampers increased use of scrap in steel production for high-end applications. Material properties not only depend on composition, but also on the production process. Composition variations can be compensated for by adapting the process. To achieve material of constant quality, excellent predictive material and process models must be available.

Using less material

Significant energy savings can be achieved by producing metals from scrap (recycling) instead of ore. However, the higher the content of scrap, the more difficult it is to maintain the required quality standards and tolerances. This hampers increased use of scrap in steel production for high-end applications. Material properties not only depend on composition, but also on the production process. Composition variations can be compensated for by adapting the process. To achieve material of constant quality, excellent predictive material and process models must be available.

Data enhanced predictive modeling

Models currently in use are usually based on mathematical descriptions of the underlying physics. Physics based models are not accurate enough for process compensation of variations in steel produced with higher scrap ratio. The absence of sufficiently accurate models is hampering the adoption of recycled steel and small batch special purpose steel.

This project focuses on creating and enhancing physical models that are qualified for baseline behavior with dataanalysis and machine learning techniques to quickly adapt predictive models to specific compositions. The availability of such models is essential for the adoption of recycled steel in high-end applications.