(617e) The Effect of Numerical Diffusion On Scalar Mixing Prediction Using CFD

Liu, M., DuPont Engineering Technology

Numerical diffusion is an intrinsic part of a numerical method for a convection-diffusion flow system. When CFD is used to predict mixing and reactions in a chemical reactor, the predicted results are often tainted by numerical diffusion since the physical diffusion, molecular or turbulent, could be overwhelmed by the numerical diffusion. This is almost guaranteed to be the case for scalar mixing in a laminar flow of industrial size. It has been believed that turbulent diffusion is several orders of magnitude larger than molecular diffusion, and therefore, turbulent mixing prediction should not be affected by numerical diffusion.  In this talk, a method is first presented to estimate the order of magnitude of numerical diffusion in a CFD solution of a continuous flow reactor. Several examples, from microfluidic mixers to industrial sized stirred tank reactors,  will be discussed and the magnitude of both physical and numerical diffusivities will be compared. Effect of numerical diffusion on mixing time predictions will also be discussed.