View article

Four different methods are compared for evaluating fluid flow velocities in the liquid pool in the mold region of a continuous caster of steel slabs. First, the instantaneous and time-averaged flow pattern in a 0.4-scale water model with single-phase flow is quantified using particle image velocimetry (PIV). Next, three-dimensional computations are performed to calculate the time-average flow pattern in the same system using a conventional finite-difference program, CFX, with the conventional K–e model for handling turbulence. Transient computations are then performed using a 1.5-million node grid to resolve the turbulent eddies, both without a turbulence model (direct numerical simulation: DNS) and with a subgrid scale model (large eddy simulation: LES). Finally, measurements are obtained using electromagnetic sensors embedded in the mold walls of an operating steel slab casting machine. The comparisons reveal remarkable quantitative agreement between all four methods for the overall time-averaged flow pattern and surface velocities for these conditions. The time-averaged K–e model is capable of accurate quantitative calculations of the steady flow field with the least effort, but has difficulty with transient behavior. The LES model predicts both steady and transient phenomena, but has severe computational cost. Water models with PIV are useful and practical tools, but are difficult to extend beyond the flow pattern to practical phenomena, such as heat transfer, solidification, surface slag entrainment, two-phase flow, and particle motion. The electromagnetic sensor has the advantage of measuring the real process, but is computed to be …