We simulated a large-scale setup for a 2000 L reactor in Siemens M-Star CFD. Agitation was scaled by tip speed, so 500 rpm in an Ambr250 became ~28 rpm in the 2000 L reactor. Aeration was 0.1 vvm. We placed three hydrodynamic scenarios side by side, and the behavior was very surprising.
Bubbles only, without stirring: pure buoyancy. The bubble column alone already drives a strong upward plume.
Stirring plus 0.1 vvm: this is the interesting one. The bubble plume completely breaks the impeller's downflow. This is not classical flooding but a gas-flow-limited mixing regime that you only see at large scale. The impeller pumps down, the gas pushes up, and the gas wins.
Stirring only: a clean downdraft, but mixing is still slow at this volume when no baffles are present.
The key point: if you scale only by tip speed, you will run into hydrodynamic phenomena that never appear in small-scale systems, like this gas-driven collapse of circulation. This is why we developed our pDS downscale reactors. They are designed to replicate actual large-scale mixing behavior, including regimes where gas interferes with the flow, instead of just matching a tip-speed number or power input.