CFD - Computational Fluid Dynamics

[Astronuc]

There are a few threads in this forum about CFD, but I thought I'd establish one for a general discussion of CFD methods and applications and tie it in with others as appropriate.

In the December '07 issue of Aerospace America in the Aerospace Sciences section is a short article on the recent developments in various subfields of fluid dynamics. The first article is about CFD Methods and Applications.

The application of hybrid Reynolds-averaged Navier-Stokes (RANS)/large-eddy simulation (LES) methods to aerodynamic flows continued to be an active area of research in the U.S. and abroad. Symposia held in London, England, and in Kerkyra, Greece, highlighted the successes and shortcomings of these techniques. In addition to the canonical separated-flow applications, the method is being used increasingly for wall-modeled LES and for zonal simulations.

Research continued in the development of techniques that improve the solution behavior at the interface between RANS and LES. Boeing and TTC Technologies independently improved and applied high-order-based hybrid RANS/LES procedures to accurately simulate the flow and fluctuating pressure fields required for jet noise prediction in a coupled nozzle/jet plume model.

BTW, I strongly recommend membership in AIAA!

 

[sayak]

Hello! I am doing Master's in CFD coding in India. we are developing a C-based general purpose code that can solve various fluid and thermal problems. No LES though. We are only implementing the 2-eqn models .

 

[jaap de vries]

Its is hard now days to come up with an "in house" code that can beat the capabilities of FLUENT. C-based sub programs or User Defined Codes (UDF's) are probably a better choice, unless you work for SANDIA. I read the other day that even at SANDIA a lot of time (up to 75%) of skilled (high paid) professionals is being "wasted" in cleaning up CAD geometries and preparing them for meshing and analysis (even when transferred as STEP files). Now that FLUENT is owned by ANSYS this might change.

For myself in the field of gas turbine combustion it is still hard to get reliable results out of LES or full combustor modeling. Partly because the fundamental physics behind the combustion process is not fully understood (reaction rates, number of intermediate, etc).

Secondly, the required output of for instance NOx of less the 15 ppm for NG or <50 ppm for diesel can be within the uncertainty of the numerical analysis.

Last but not least is the fact that even a modest reaction model for methane (GRI 3.0) includes over 350 individual reactions and this combined with DNS makes problems too large for even the most powerful clusters.

I think numerical analysis should actually have its own place as a topic within physicforums since it is so broad and yet so dynamic that forums like these are ideal for fast change of ideas.