Direct reduction of iron ores: From a porous solid model to CFD simulations

Abstract

With an average annual production of 45 million tons, Germany is one of the world's largest steel producers. On the reverse side, the iron and steel industry constitutes the major industrial source of CO2. The emissions related to this industry in Germany amounted to 37.9 million tons CO2-eq in 2018. Two-thirds of the iron is produced in blast furnaces, using coke and coal to reduce iron ores. An alternative iron-making route, the Direct Reduction (DR) of iron ores, is gaining interest. The reducing gas typically used is natural gas, but syngas from biogenic resources or pure hydrogen can also be used. The direct reduction with green hydrogen can reduce the CO2 emission by 95 % compared to the blast furnace process. The conversion rate of single iron ore pellets to iron in DR processes has been extensively investigated. Simple conversion models can deliver transport and reaction parameters, which can further be used in Computational Fluid Dynamics (CFD) simulations of industrial-scale reactors for design or optimization purposes. However, this transition from a base-model to CFD cases is not apparent. On the one hand, because of the range of availability of the base-model. External and internal transport limitations are often lumped in the reaction term. This explains the disparateness of reaction rate parameters proposed in the literature. On the other hand, because of fundamental differences. The traditional shrinking core model assumes a moving reaction front separating an unreacted core and a fully converted region. This assumption cannot be properly declined for finite-volume methods, which relies on the discretization of the domain. The present study will focus on achieving a reliable transition from a porous solid model (i.e., the base-model) to CFD simulations. To this end, both the base-model and the CFD-model will be applied to the same experimental data sets describing the reduction of single pellets. The base-model has a low computational cost and can therefore be used to derive kinetic and transport data, which is further usable in the CFD-model. The CFD simulations solve for the whole computational domain, and thus, can in turn deliver accurate boundary conditions for the base-model. It will be verified that identical input parameters lead to almost identical results in both, the base-model and the CFD-model. The possible sources of discrepancies will be investigated, such as deviations in the symmetry hypothesis.