The physical model of a ten-strand billet caster tundish was established to study the effects of various flow control devices on the melt flow. Before and after the optimization of the melt flow, the inclusion removal in the tundish was evaluated by plant trials. The physical modeling results show that when combined with a baffle, the turbulence inhibitor, instead of the impact pad, can significantly improve the melt flow. A turbulence inhibitor with a longer length of inner cavity and without an extending lip at the top of the sidewall seems to be efficient in the improvement of the melt flow. Various types and designs of baffles all influence the flow characteristics significantly. The “V” type baffles are better than the straight baffles for flow control. The “V” type baffle with four inclined holes at the sidewall away from the stopper rods is better in melt flow control than the one with one inclined hole at each sidewall. The combination of a well-designed turbulence inhibitor and an appropriate baffle shows high efficiency on improving the melt flow and an optimal proposal was presented. Plant trials indicate that, compared with the original tundish configuration in prototype, the inclusions reduce by 42% and the inclusion distribution of individual strands is more similar with the optimal one. The optimal tundish configuration effectively improves the melt flow in the ten-strand billet caster tundish.

The flow characteristics and inclusion removal in a ten-strand continuous casting tundish were investigated with physical modelling and industrial trials. The results show that, among the strands, the strand with the minimum dimensionless time of the first appearance of tracer at the tundish exit appears to be the worst one for inclusion removal, while the strand with the maximum dimensionless mean residence time shows the best inclusion removal efficiency. The inclusion number decreases with increasing inclusion size for all strands. The inclusion number distribution among strands is the same for all inclusion sizes and the descending order of inclusion number is basically consistent with the ascending order of dimensionless mean residence time among individual strands. However, when the strand with the minimum dimensionless time of the first appearance of tracer at the tundish exit is not the same one with the minimum dimensionless mean residence time, the former seems to be inferior to the latter for inclusion removal.

Combining with the physical model of level fluctuation in a thick slab continuous casting mold with the cross-section of 1500 mm×280 mm and argon blowing, the rationalities of estimating the level fluctuation by three traditional quantitative approaches were discussed, and the effects of gas flowrate, casting speed, and the immersion depth of submerged entry nozzle (SEN) on the level fluctuation were also investigated. As a result, it seems that three traditional quantitative approaches are not very suitable for estimating the level fluctuation in a mold with argon blowing, so a new approach for estimating level fluctuation in the mold with argon blowing was presented. The experimental results show that the level fluctuation is mainly in the region around the nozzle wall. When the casting speeds are larger than a certain value, there is the escape of large bubbles near the nozzle wall, which causes an obvious increase of level fluctuation. Furthermore, optimal process parameters, viz., the gas flowrate of 6 NL/min, the casting speed of 1.1 m/min, and the immersion depth of 170 mm, are presented to restrain the level fluctuation by a physical model.