No abstract

No abstract

The liquid steel flow and the particle transport of non-metallic inclusions in the liquid domain inside the strand of a continuous steel casting system are numerically investigated. The whole liquid domain inside the solidified shell is simulated, and the loss of liquid steel due to solidification is considered. Assuming a constant drift velocity, a kind of Euler-Euler multiphase model is implemented to calculate the inclusion behaviour inside the liquid steel, respectively the inclusion removal at the boundaries. The turbulent flow of the liquid steel and the inclusion behaviour are calculated. The resulting distribution of the inclusions in the slab is compared with measurements. The influence of inclusion size (0 to 1 mm) and casting speed on the inclusion distribution is considered.

The Separation of nonmetallic particles by buoyancy in tundishes is investigated theoretically. It is shown, that the best possible separation is achieved it the projection of the region through which flow from the inlet to the outlet mainly occurs, covers as large a part of the surface as possible. In contrast, the volume of this region, its distance from the surface and the mean retention time do not influence separation directly. Numerical calculations of the flow fields and particle separation in two different tundish geometries with and without a baffle predict no significant differences in the capability of separating nonmetallic particles. This surprising result may be explained by the principle stated above. In contrast to these findings, observations made during production provide indirect evidence that the insertion of a baffle reduces the concentration of Al2O3 at the outlet of the tundish significantly. This would mean that besides buoyancy, another process contributes significantly to the separation of nonmetallics. Such a process is not identified in this paper, but it is shown that turbulent diffusion and inertial impaction are not likely candidates.

The removal of non-metallic inclusions due to buoyancy forces in tundishes in continuous casting systems is considered. The maximal theoretical removal rate is determined by the flow rate through the tundish, the magnitude of the tundish fluid surface area and the particle terminal rising velocity, depending on the particle size. Two reasons, why the particle separation is worse than the maximal possible one are an unsuitable fluid flow pattern and the turbulent particle diffusion. An analytical discussion of simple flow patterns (parallel flow with different velocity profiles, dead regions, swirling flows) and diffusion shows how they influence the particle removal. Using these simple considerations, it is demonstrated that the often used RTD (residence time distribution)-curves are inappropriate to estimate the particle separation in tundishes; only the direct measurement and CFD-simulation of particle removal should be used. The common CFD-methods are affected with numerical errors such as numerical diffusion for particle concentration simulations with finite volume methods and interpolation errors for particle trajectory calculations. These errors influence significantly the calculated particle removal curves; they are nonsystematic and difficult to quantify. Neue Aspekte zur Partikelabscheidung in Verteilern. Das Abscheidungsverhalten von nichtmetallischen Partikeln in StranggieBverteilern infolge von Auftrieb wird untersucht. Zunachst wird die maximal moqllchs Abscheidung bestimmt, die sehr einfach aus Durchsatz, Verteileroberflache und Partikeldriftgeschwindigkeit (abhanqlq von der PartikelgroBe) bestimmt werden kann. Dann werden zwei Ursachen untersucht, die tur eine nicht optimale Abscheidung verantwortlich sind: ungOnstige Stromungsform und turbulente Partikeldiffusion. Es wird mit analytischen Methoden aufgezeigt, wie sich verschiedene einfache Grundstromungsformen (Parallelstromunq mit verschiedenen Stromungsprofilen, Totraurne, Drallstromungen) und Diffusion auf das Abscheideverhalten auswirken. Anhand dieser einfachen Beispiele kann gezeigt werden, daB die oft verwendeten Verweilzeitkurven im allgemeinen ungeeignet sind, urn das Abscheideverhalten eines Verteilers zu beurteilen. Statt dessen bleibt nur die direkte Messung oder Simulation von Abscheidevorgangen. Bei der Oblichen CFD-Simulation ist allerdings Vorsicht geboten: Die numerischen Fehler der verschiedenen Verfahren (numerische Diffusion bei finiten Volumen, Interpolationsfehler bei Partikeltrajektorien) vertalschen das Ergebnis deutlich und stellen einen unsystematischen und schwer quantifizierbaren Fehler in den Abscheidekurven dar.