Publication details

I. Matino, M. Vannucci, A. Petrucciani, A. Zaccara, V. Colla, T. Cucinotta, M. Ferrer Prieto. "Advanced Data-Driven Methods for Predicting Electric Steelmaking Slags," 9th International Slag Valorisation Symposium (iSLAG 2025), April 8-11, 2025, Leuven, Belgium.

Abstract

Circular economy and industrial symbiosis are strategic pillars for European growth, economy, sustainability, and competitiveness, as emphasised in the Strategic Research Agenda of the European Steel Technology Platform (ESTEP) and of the Clean Steel Partnership. In this background, by-products are acquiring ever more importance, and the steelmaking industry (where about 90% in mass of the by-products are slags) is investigating solutions allowing their optimal management in terms of reuse and recycling. In electric steelmaking, slags are produced in Electric Arc Furnaces (EAF) and in Ladle Furnaces (LF), and sometimes their management is not optimised since their features (e.g. their composition) are not continuously monitored. Knowing in real-time the slag composition with standard analytical techniques is challenging. The procedure often requires long times and presents difficulties connected to harsh environment (for liquid slag) and high heterogeneity of slags (especially at solid state), that would require costly and lengthy multiple sampling to ensure representativeness. The sample preparation, whose complexity depends on the adopted analytical method, also affects measurements accuracy, and sometimes only semi-quantitative methods are used. On the other hand, a more precise and anticipated knowledge of slags composition provides opportunities from the point of view of both process optimisation and slag valorisation. Slag conditioning and metallurgical processes control, as well as slag handling, can be improved using information on slag features, by enabling implementation of suitable practices for slag treatment, recycling and valorisation. Therefore, besides the requirements of new sensing devices, digital tools can be used to estimate slags composition. Different models can be found in the literature on the electric steelmaking process but they are generally not focused only on the slags simulations and/or are too complex for real-time use in decision support systems finalised to the slag monitoring and management. For this reason, among the different activities, during the European RFCS project "iSlag", different models were developed for both offline analyses (e.g. a flowsheet model considering the whole electric steelmaking route10, and allowing computing, among others, steel and slags amounts, compositions and temperature, electric energy requirments, etc.) and online estimate of slag compositions. For this last goal, data-driven models were developed in their pure or hybrid configurations, by employing techniques based on Artificial Intelligence and particularly on Deep Neural Networks also combined with phyisics-based model. This paper focuses on them.

Download paper

Last updated on
11 April 2025