HAlMan Workshop

Decarbonization Strategies and Relevant AI/ML Solutions for the Metallurgical Industry

About Workshop

This workshop explores innovative approaches to decarbonising manganese (Mn) pre-reduction processes by combining advances in process development with state-of-the-art Machine Learning (ML) methods. HAlMan consortium partners NTNU and NTUA will present their work on process development and decarbonisation strategies, while MET4 and SINTEF will showcase AI/ML models and optimization frameworks for process modelling and optimization. Together, the presentations will demonstrate how integrating metallurgical expertise with data-driven approaches can support more sustainable and efficient manganese production.

Presentations & Speakers

lorem ipsum

Prof. Jafar Safarian

Norwegian University of Science and Technology (NTNU) | Project HAlMan coordinator

Jafar Safarian is a professor in extractive and process metallurgy with over 32 years education and experience in this field. He got his PhD in 2007 at NTNU on “Ferromanganese slags reduction by selected forms of carbonaceous materials”, and his PhD study was followed with about 5 years academic and research activities as a PostDoc fellow/researcher at NTNU. He worked more than three years at SINTEF Materials and Chemistry during 2012-2015, and again joined NTNU as a fulltime employee from 2016 with Associate Professor position, and full Professor position since 2021. He has academic teaching duties, and does currently research on solar grade silicon production and recycling, sustainable alumina production, hydrogen reduction processes, manganese dioxide production, valorisation of waste, etc. He has supervised several PhD and PostDoc researchers as well as master’s and bachelor’s students. He has management duties in several educational and research projects such as HydroMetEC, supported by EIT Raw Materials, FME-SUSOLTECH supported by the Research Council of Norway, and HALMan, HARARE, and ENSUREAL, supported by the Horizon Europe 2020 program.

Translating Experiments into Data: Towards Intelligent Control of Carbonation for Alumina Recovery

Carbonation of sodium aluminate/sodium carbonate solutions involves complex chemical equilibria and competing precipitation pathways, making process control difficult using a single process parameter such as pH. The formation of unwanted phases such as dawsonite highlights the need for a more comprehensive approach to identifying the carbonation endpoint. This work combines thermodynamic modelling with systematic monitoring of multiple solution properties during carbonation to link chemical behaviour with experimentally measurable process signals. The approach aims to reduce the experimental burden, improve process understanding, and generate reliable data for future data-driven modelling and intelligent process control.

Danai Marinos

PhD candidate, Supervising Researcher / Hydrometallurgical Purification Processes

Danai Marinos graduated from the School of Mining and Metallurgical Engineering at the National Technical University of Athens (NTUA) in 2013 and obtained her MSc in Extractive Metallurgy from the Colorado School of Mines in 2014. Her research has focused on hydrometallurgical processes, including the beneficiation and recycling of spent lithium-ion batteries and the precipitation of aluminium hydroxide from sodium aluminate solutions using CO₂. Since returning to NTUA in 2014, she has been a researcher with the Technologies for Sustainable Metallurgy group and has contributed to several European research projects. Her expertise includes the leaching of primary and secondary raw materials, precipitation, ion exchange, solvent extraction, thermodynamic analysis, and acid recovery from complex hydrometallurgical systems.

Learn, Constrain, Decide: Two AI/ML Case Studies in Pyro- and Hydrometallurgy

Two metallurgical process units, one recipe: encode the physical chemistry into the features or into the output constraints, fit an AI/ML surrogate under a validation protocol matched to the data, and optimise down to a handful of operating points an engineer can act on. The cases are hydrogen pre-reduction of manganese ores (JMAK kinetics, gradient boosting, NSGA-II) and carbonation endpoint control for alumina recovery (a Gaussian-process soft sensor with chemically consistent outputs).

Michail Mavroforakis, PhD

Chief Technology Officer (CTO)

Michail Mavroforakis, is Chief Technology Officer at MET4. He specializes in applying AI/ML methods to optimize physicochemical processes. A Senior Member of the IEEE, his AI/ML research was recognized with the IEEE Best Paper and Outstanding PhD Dissertation Awards, and the EURASIP Best PhD Award. He previously served as Research Assistant Professor at the University of Houston, USA, and as the Group CISO and the Chief Data Officer at the National Bank of Greece.

From Kinetics to Surrogates: Modelling Hydrogen Reduction of Manganese Ores for Decarbonised Mn-Alloy Production

This presentation will show how physics-based kinetic modelling and machine-learning surrogate models together can accelerate the development and optimisation of metallurgical processes. Examples from the HAlMan project — isothermal and non-isothermal modelling of hydrogen reduction of manganese ores, and neural-network surrogates trained on the resulting data — will illustrate the integration of process knowledge with AI tools to support decarbonisation, improve resource efficiency, and enable rapid evaluation of operating conditions for manganese pre-reduction and metallothermic production.

Kai Tang

Senior Research Scientist, SINTEF Industry

Kai Tang is a Senior Research Scientist at SINTEF Industry in Trondheim, Norway, where he has worked since 2002. His research centres on the thermodynamic and kinetic modelling of high-temperature metallurgical processes — CALPHAD-type modelling of slags and alloys, silicon and ferroalloy production, recycling of critical raw materials — and, increasingly, on machine-learning methods for process modelling and optimisation. He has led or coordinated numerous national and EU projects, including the ongoing HAlMan and APOLLO projects, and has published over 200 papers (h-index 31, over 3000 citations).