Industrial Reactor Evaluation with Uncertainties In Mixing Performance and Reaction Kinetics

Frequently the Fluid Mechanics & Mixing Discipline of The Dow Chemical Company (Dow) is asked to assess the mixing-sensitivity of industrial chemistries.  The chemistries may be undergoing scale-up or are in some way experiencing process difficulties.  Sometimes the aim is to minimize impurities of any sort.  In other, perhaps more challenging, cases the aim is to maintain a constant distribution of by-products to ensure that the product performance is unchanged.  Most desirably, we have access to a detailed intrinsic kinetic rate model and detailed knowledge of the fluid dynamics, so that the pertinent time scales can be calculated and a Damkohler analysis completed for mixing sensitivity.  Reality is frequently quite different. 

Generally, we have the following scenarios with respect to kinetics:

-          A kinetic model exists or can be derived from the correct type of existing lab or process data.

-          A kinetic model does not exist, but a reactor mole balance & product distribution is available or can be derived from process data.  Reactor sampling may be impossible, meaning the engineer must resort to approximations based on plant feed and product flows, compositions, tank level changes, etc.  Calorimetric data may exist.  If resources allow, a kinetic model can be derived from molecular modeling and thermochemical kinetics, then validated with product data.

-          No intrinsic kinetic data exist.  Previous process data was taken with a transport limitation.

And generally we have these scenarios for flow field:

-          Vessel geometry is known precisely and the flow field is well-characterized.

-          Vessel geometry is known but flow characteristics are not known.  If the reaction is localized, process instrumentation such as thermowells help understand some spatial reaction characteristics.

-          Studies in lab glassware, such as a round-bottomed flask, have given satisfactory performance.  Blending is inherently rapid.  No significant changes to the flow field have been studied.

Furthermore, there can of course be large interactions between the turbulence and reactions. 

The next scale often must fit in to an already-existing reactor somewhere.  In this case, the stakeholders need assurance that the existing reactor can work, and whether any changes must be made to the agitation, feed introduction, etc. to ensure success.  Sometimes the reactor is yet to be built, so that many degrees of freedom for design are available.

This talk will focus on real examples of recent industrial problems encountered and the methodologies used to address them.