Portal de Periódicos da CAPES

Data-assisted CFD modeling of transient blast furnace tapping with a dynamic deadman

2019; Elsevier BV; Volume: 73; Linguagem: Inglês

10.1016/j.apm.2019.04.024

1872-8480

M. Vångö, Christoph Feilmayr, Stefan Pirker, T. Lichtenegger,

Rock Mechanics and Modeling

The blast furnace (BF) hearth’s condition is crucial for a long and healthy campaign life. Previous numerical investigations on BF drainage commonly treated the deadman as a static, porous medium. In this work, we employ a coupled computational fluid dynamics (CFD) – discrete element method (DEM) to account for its dynamic behavior. The model is utilized in several short-term simulations to generate a database of deadman states based on iron and slag levels, coke properties and the burden weight distribution. In contrast to the previous Eulerian deadman descriptions, we directly obtain an inhomogeneous description of the porosity from a particle size distribution. The database is utilized in the dynamic void fraction model, to describe the transient deadman in a fully Eulerian framework for long-term simulations. In essence, we present a strategy to mitigate the coupling between the dynamics of a packed bed and the fluid flow through it by connecting the bed’s evolution to global state parameters such as the levels of liquids in which the particles float. This methodology is not restricted to blast furnaces, but can also be applied to other problems involving slow granular motion with dynamic interstitial fluids. The model is first validated with experiments of water drainage through a lab-scale, particle-filled tank, and is shown to give similar results as CFD–DEM for both sitting and floating particle beds. We then demonstrate our model’s advantages and necessity on isothermal simulations of a full-scale BF hearth for various deadman setups. We further highlight the effects of a coke free region due to a floating deadman as well as an almost impermeable dense center on the flow. Finally, by comparing simulations of a dynamic and a fixed floating deadman, we show that the dynamic behavior significantly affects the iron and slag drainage rate, which results in a 20 min extended duration after 120 min tapping. To conclude, we present a highly performing approach to study the behavior of fluid flow through moving granular beds without costly contact detection but with information obtained from previous DEM simulations.