(270a) Power Dissipation and Power Numbers for a Retreat-Blade Impeller in a Torispherical-Bottomed Pharmaceutical Reactor Vessel Under Different Baffling Configurations

In the pharmaceutical industry, glass-lined reactors and vessels are often utilized to carry out a variety of different unit operations and especially reactions.  Within these systems, the vessel impeller, and baffle(s) are typically glass-lined in order to provide superior corrosion resistance, prevent product contamination, and enhance cleanability.  Because of the fabrication issues associated with glass lining metal surfaces, including baffles and impellers, these reactors are typically equipped with a single retreat-blade impeller (RBI) placed close to the tank bottom and a single baffle mounted from the top.  Despite their common use in industry, these systems have received relatively little attention.

In this study, the power dissipation and the Power Number, Po, were experimentally obtained and computationally predicted in a 60-L scale-down version of a full-scale reactor for pharmaceutical production provided with a torispherical bottom, a RBI, and a single beavertail baffle, in order to simulate a geometrically similar production reactor.  In addition, different baffling configurations were studied, i.e., fully baffled and unbaffled systems.  Water and sucrose solutions of different viscosity were used to examine a wide range of Reynolds Numbers (Re).  At lower Re values (Re~100) all systems produced similar Po values irrespective of baffling.  However, as Re increased, Po for the unbaffled and, to a more limited extent, unbaffled systems decreased appreciably with respect to the Po for the fully baffled system.  Po was found to be constant at 0.73 for the fully baffled system at high Re. For the unbaffled system Po decreased with Re over the entire Re range explored here (100<Re<300,000), asymptotically approaching the 0.2-0.25 value.  Po for the partially baffled system was intermediate between the other two cases, with Po approaching the ~0.5 value at high Re.

Computational Fluid Dynamics (CFD) was used to model all the above-mentioned systems, but only at high Re, using different modeling approaches including single reference frame (SRF), multiple reference frame (MRF), and sliding mesh (SM) models.  The k-ε turbulence model was used in all cases.  The CFD simulations resulted in the prediction of the power dissipated by the impeller and therefore Po.  These predictions were compared with the experimental results.  In the fully baffled system, the predicted Po values using MRF modeling were in close agreement with the experimental results.  In the partially baffled system, the results obtained with MRF modeling were very consistent with the experimental results.  However, even better agreement was obtained when the much more computationally expensive SM modeling was used.  Finally, the simpler SRF approach proved to be very appropriate to model the unbaffled system, and good agreement between the simulation predictions and the experimental results was obtained, but only if the surface deformation of the liquid-air interface typically observed in unbaffled systems was small. 

Finally, the Po-vs.-Re experimental data were additionally fitted with a semiempirical equation whose parameters were adjusted for each baffling system.  The resulting equations closely matched the experimental data.