CFD Online Discussion Forums

CFD Online Discussion Forums (http://www.cfd-online.com/Forums/)
-   CFX (http://www.cfd-online.com/Forums/cfx/)
-   -   advection term discretization (http://www.cfd-online.com/Forums/cfx/24183-advection-term-discretization.html)

N.R. June 26, 2007 17:27

advection term discretization
 
Hi,

I would like to use a 2nd order central difference scheme for discretization of the advective term. (laminar flow, steady state). Any clues, why this doesn't work?

(Even something like including this in the CCL

ADVECTION SCHEME:

Option = Central Difference

END

will not work)

Numerical stability is not a problem. Accuracy is very important... What could I use as a substitute, if CFX doesn't offer that feature?

Thanks in advance,

N.R.

Robin June 26, 2007 17:37

Re: advection term discretization
 
This should work. Pre will raise an error, but the solver will still run.

-Robin

N.R. June 26, 2007 18:10

Re: advection term discretization
 
Thanks Robin,

How can you see that CFX is actually running the 2nd order central differencing scheme and not switching to the standard "high resolution" scheme? Just from residual data?

When I use the QUICK solver, which is also not in the GUI, it will not give an error message.

Thanks,

N.R.

Robin June 27, 2007 11:45

Re: advection term discretization
 
Check the CCL in your OUT file.

Karke June 29, 2007 00:27

Re: advection term discretization
 
Hi,N.R. and Robin:

As you know, there are many features that did not appear in the GUI, So how can i know much more these features and usage?

Thanks a lot!


Karke June 29, 2007 03:01

Re: advection term discretization
 
Central Difference This option is available when using the large eddy simulation turbulence model and is recommended for these cases.


mrkmrk December 8, 2014 07:26

Hi,

I have a question about ANSYS CFX discretization scheme of the Advection term,
CFX uses this relation for calculation of variable "phi" value in advection term,
phi_ip=phi_upwind+(betta*grad(phi)*deta_r)

where "betta" is the Blend Factor and "r" is the vector from the upwind node to the ip (shown in the attached fig)

In the central difference scheme "grad(phi)" is set to the local element gradient and in the "Specified Blend Factor" scheme "grad(phi)" is set to the average of the adjacent nodal gradients,
Can anyone tell me what is the meaning of "local element gradient" and "average of the adjacent nodal gradients" in the attached grid network that is shown below,
http://oi57.tinypic.com/34xhrpl.jpg
http://tinypic.com/view.php?pic=34xhrpl&s=8

adunne304 December 8, 2014 14:45

local element gradient is explained a few pages after where you've grabbed that image from in the theory guide. It's the variable gradient at the node (n1, n2 or n3). I.e. single values are stored at the node. For some terms in the discretized equations, gradients are required at these nodes.

Average of the adjacent nodal gradients is where at the face of the 'computational cell' (or ip) you compute the average of the nodal gradients around it (average of gradients at n1, n2 and n3)

mrkmrk December 8, 2014 15:22

Hi Adrian,

thanks for your help,

and how this gradient is calculated regarding to the flow direction? (for simplification I attached another fig)

As you know we are not allowed to use the central difference discretization scheme for this term then what is the second order backward expression for ip1?
http://oi57.tinypic.com/2qicyvs.jpg


All times are GMT -4. The time now is 05:59.