Study of computational issues in simulation of transient flow in continuous casting

Unsteady three-dimensional turbulent flow and heat transport in the liquid pool during continuous casting of steel slabs has been computed using several different computational models, domains, grids, and inlet conditions. The most advanced computations employ a large-eddy simulation code, UIFLOW with a second-order central-differencing scheme, 1.6 million nodes and a realistic simulation domain including the complete submerged entry nozzle. The model has been validated in previous work through comparison with PIV measurements in caster water models, and with velocity and temperature measurements in an operating steel caster. The present computations are compared with flow measurements in a full-scale water model and with heat flux measurements in a jet impingement test problem. Results are compared between model domains of the full caster with symmetric half-caster and two-fold symmetric quarter-caster simulations. The effects of thermal buoyancy and the solidifying steel shell walls are studied independently. The effect of different inlet conditions is investigated by comparing results including nozzle simulations that are both coupled and uncoupled with the mold domain and a simplified nozzle geometry. The importance of the Sub-grid scale (SGS) model for treating the small turbulent eddies is investigated through simulations with and without the Horiuti SGS K model. A rigorous grid refinement study is undertaken, which indicates criteria for choosing the element size near the walls. Accurate heat transfer predictions are more difficult to attain than accurate velocities. Finally, comparisons are made with Reynolds-averaged approaches, including standard K-ε and low Re-number K-ε model computations of the same system. The relative advantages and disadvantages of the different flow simulation methods are evaluated.