AI-based monitoring of steel scrap properties for improved utilization in scrap based steelmaking

Scrap-based steelmaking in the Electric Arc Furnace (EAF) allows to produce high quality steel grades on the basis of secondary raw materials. However, due to the fluctuation and quality impairment a tight control of the properties of steel scrap is required. In the framework of the EU-funded project “Artificial Intelligence Technologies for European Process Industry digital transformation (s-X-AIPI)”, an intelligent, AI-based system for continuous monitoring of scrap properties, directly coupled with a scrap mix optimisation for steel qualities to be produced was developed and implemented at the Sidenor Electric steelmaking plant in Basauri, Spain. The system is designed to enhance the selection process of the charge materials for EAF steelmaking through the integration of a self-managing AI framework. Predictive machine learning models are used to improve the reliability and precision for characterisation of scrap properties like metallic yield, composition and specific energy requirements from historical and most recent production data, to provide upto-date information to a scrap mix optimization calculation. When significant deviations between predicted and actual liquid steel composition occur, an autonomous event handler initiates a retraining of both the prediction models and the scrap characterisation tools, to adjust the scrap property parameters, the predictive AI models, and/or the scrap mix optimizer constraints. Supported by a knowledge base, this AI framework ensures consistent and reliable scrap management under varying and changing process conditions.