CFD Online Discussion Forums

CFD Online Discussion Forums (http://www.cfd-online.com/Forums/)
-   CFX (http://www.cfd-online.com/Forums/cfx/)
-   -   Question about CFX pressure averaging in a surface (http://www.cfd-online.com/Forums/cfx/116995-question-about-cfx-pressure-averaging-surface.html)

Anna Tian April 30, 2013 07:38

Question about CFX pressure averaging in a surface
 
Hi,

If there is a strong vortex across a surface and we'd like to extract the averaging pressure and total pressure value of this surface, could CFX output the averaged value accurately for this case? If not, how to improve the accuracy?

oj.bulmer April 30, 2013 08:00

If your solution is mesh independent and converged, then it should. I don't still understand what doubt prompted you to ask your question.

OJ

JuPa April 30, 2013 08:56

For your info:

Be careful with averaging if your simulation is transient. You should only turn the transient statistics on (transient average, transient standard deviation etc) after the initial transient has diffused. I.e. let your monitor points of interest (such as temperature or velocity) settle before recording transient averaged results.

If your simulation is steady state then ignore what I just said.

Anna Tian April 30, 2013 10:33

Quote:

Originally Posted by oj.bulmer (Post 424153)
If your solution is mesh independent and converged, then it should. I don't still understand what doubt prompted you to ask your question.

OJ


Because if there is a vortex cross that surface, we will have large pressure or total pressure gradient on that surface. If we do averaging, we will need to approximate the value for the area between the cell centers. That could introduce averaging errors, right?

Even we have grids independence, but that error could still be not small, right?

oj.bulmer May 1, 2013 11:04

I see. If you have a look at the definition of area averaged values:

\phi_{area-average} = \frac{1}{A} \int \phi dA = \frac{1}{A} \sum_{i=1}^{n} \phi_i \vert A_i \vert

It is evident that averaging will suffer if the value of Ai, ie facet value of a single cell, is unreasonably (too) large to represent a distinct value of \phi_i over its area, especially at the regions of high gradients. I think smaller the cell size at these critical areas, better would be your averaged values.

I would use adaptive meshing (based on gradients) If I were too concerned with region-averaged statistics, while obtaining the mesh independence. And this distribution can be further refined for more accuracy.

OJ


All times are GMT -4. The time now is 22:27.