Case Study: OCAS improves steel property prediction with Intellegens Alchemite™

Understanding and improving steel performance at OCAS

OCAS, the steel research centre jointly owned by ArcelorMittal and the Flemish regional government, wanted to better understand the complex Processing-Structure-Property relationships that drive steel performance. Working with Intellegens, it set out to test whether machine learning could handle real-world, sparse experimental data and whether microstructural image data could improve prediction of steel properties while reducing the need for costly physical testing.

Intellegens implemented its Alchemite™ deep learning software to model OCAS steel data, combining processing variables, material chemistry, and microstructure images. The result was high predictive accuracy, with R² values above 0.8, improved performance when image data was included, and evidence that some image data could be reduced by up to 30% without harming accuracy. Intellegens also helped identify important inputs through explainable AI tools and generated processing parameter combinations that met target steel property requirements.

OCAS

Lode Duprez

Chief Scientific Officer