[meroney]

Fluent provides an option to solve various turbulence models while simultaneously specifying non-Newtonian fluid properties (Eg. power law, Carreau, Cross, Herschel-Bulkley, etc.). This model has recently been used to calculate flow around mixers in anaerobic digesters in the reviewed literature. I would like to convince myself that this is a viable model; hence, I have attempted to validate the protocol by reproducing various experimental and DNS data for non-Newtonian power law fluids flowing in pipes. The experimentalists report velocity profiles, turbulence components, pressure drops, shear, etc.

First, I succesfully reproduced Laufer's turbulent pipe flow Newtonian fluid data using the RMS model to convince myself I was using Fluent properly. Results were VERY good in all respects. Then I have attempted to replicate the DNS study by Rudman et.al. (2004) J. Non-Newtonian Fluid Mech, 118, pp 33-48 where they predicted velocities and turbulence using DNS and compared to shear-thinning CMC solution data. I used their specifications for viscous properties, RMS and ke-realizable models, very fine grids near the wall (similar to earlier Laufer simulation...ie. y+~O(1)), and iterated to residuals below 10^-4.

The results produced too high pressure drops, too large wall shears, and inappropriate u', v', and w' behavior. I then attempted to replicate the laser-doppler measurements of Pinho and Whitelaw (1990), J. Non-Newtonian Fluid Mech, 34, 129-144. Results were similarly dismal.

Has anyone had a better experience? Can you share details? Can you suggest what my problem might be if you have had better success?

Thanks.