The introduction of the appropriate size and precise location of flow control devices such as dam, turbulence inhibitor, etc., helps to modify the flow pattern and minimizing short circuiting and dead zone. Beside this, these also create the surface directed flow and maximize the residence time available for the flotation of inclusions and assimilation of the reaction products from the molten steel into the slag phase. These can be products of deoxidation, reoxidation, precipitation, emulsification and/or entrainment of refractory components into the melt and thus encompass both indigenous and exogenous inclusions. To this end, both the numerical and physical simulations were carried out mainly for three cases: a) in absence of flow control devices (i.e., bare tundish), b) in the presence of a dam, and c) with the application of turbulence inhibiting device (TID) and dam combination in the existing tundish configuration. The commercial CFD (computational fluid dynamics) package FLUENT ® was used to predict the flow field prevalent in the water model tundish at steady state, whereas in the experimental program, both Particle Image Velocimetry (PIV) techniques for flow measurements and tracer dispersion experiments for concentration measurements were applied in the present study. Among all types of configurations applied in the present study, a combination of TID with holesϩa dam work reasonably found to be an optimum configuration of the four-strand tundish regarding inclusion floatation. A superior strand similarity is also achieved in this configuration. Also the predicted time averaged horizontal and vertical components agreed withinϮ10 % with the experimentally derived ones.

A summary of computational work reported in the literature on tundish hydrodynamics has been presented wherein, it is shown that a diverse range of both computational (e.g., nodal configurations, boundary conditions, inlet turbulence etc.) and physical parameters (e.g., size, number of strands, inlet mass flow rates etc.) were applied. Accordingly, the conclusions drawn were found to vary from one study to another. In the present work, an attempt has therefore been made to assess computationally the role of various mathematical model parameters. To this end, mathematical model results were validated against experimental measurements on Residence Time Distribution (RTD) parameters derived from water model tundish. Experimental measurements of RTD were carried out continuously by monitoring conductivity of water at the tundish exit port on a digital computer using a DAS interface. On the other hand, numerical calculations were carried out via the commercial CFD (Computational Fluid Dynamics) software FLUENT, 6.0. The combined experimental and computational study indicated that a sufficiently small grid resolution (control volume of the order of 10 Ϫ6 m 3) is necessary to arrive at a practical grid independent solution. Furthermore, Reynolds stress model was found to simulate RTD in the system somewhat superior to the standard coefficient k-e model. Through comparison of the predicted results with experimental measurements, a set of optimal mathematical model configurations was deduced. It was demonstrated that mathematical model configured in this work is sufficiently reliable and robust as this leads to estimates of RTD parameters (e.g., t min , t max , t mean) close to experimental measurements in a tundish with and without flow modifiers.

No abstract

No abstract

An exhaustive literature search indicates that, despite a large number of physical and mathematical model studies, very little efforts have been made to assess predicted flow and turbulence parameters in the tundish directly against equivalent experimental measurements until recently. Consequently, experimental measurements on the instantaneous velocity and residence-time distribution (RTD) were carried out in a scaled water model of a four-strand billet-casting tundish. While particle-image velocimetry (PIV) was applied to measure instantaneous flow, the electrical-conductivity measurement technique was applied to determine the RTD. Through PIV, the mean and the fluctuating components were derived along the central vertical plane of the tundish at two different liquid inflow rates: 1.55 ϫ 10 Ϫ4 m 3 /s and 3.10 ϫ 10 Ϫ4 m 3 /s, respectively. Similarly, RTD curves were obtained for tundish operations without and with a dam ϩ turbulence inhibitor device (TID). Parallel to these operations, the flow and tracer dispersion were numerically predicted by FLUENT ® . It is shown that the predicted time-average velocity components within the bath bear excellent correspondence with PIV measurements. On the assumption of isotropic fluctuations, turbulent kinetic energy was derived from experimental measurements, which agreed moderately with predictions. Furthermore, the experimentally derived fluctuating velocity components were compared with those obtained from the Reynolds stress model. This indicated very reasonable agreement between measurement and predictions (within Ϯ20 pct). Despite such a difference, however, the extent of agreement between the measured and computed C curves was found to be excellent.