cell-based data to node-based data

gonski · March 31, 2009, 18:37

Hi all,
My CFD code provides fluid velocity on cell center. But I need velocity on grid nodes for some special calculation. So, which scheme is the best for me to interpolate the cell-based data to grid node?
Your hints would be greatly appreciated.
Cheers,
Gonski

gory · March 31, 2009, 20:49

Volume-weighted average, Least-squares fit (possibly with weights), etc. Or generate a dual grid whose nodes coincide to the cell centers of the original grid?

gonski · March 31, 2009, 21:08

Now I try to use gradient interpolation, i.e. interpolate velocity from center of cells around the considered node to the node, then average it. the gradient is calculated by a leas-square or guass method. I guess if gradient is based on least-square method, it should be least-square fit mentioned by you? I found this method produce strange results on boundaries.

RenardP · March 31, 2009, 21:58

You can search on NASA reports site after Neal T Frink, he presents a simple method for interpolating the cell centered values to the grid nodes. His method works very well on unstructured grids.

gonski · March 31, 2009, 22:02

Quote:

Originally Posted by RenardP

You can search on NASA reports site after Neal T Frink, he presents a simple method for interpolating the cell centered values to the grid nodes. His method works very well on unstructured grids.

Will you please tell me which report published by Neal T Frink?
Cheers,
Gonski

RenardP · April 1, 2009, 00:10

Tetrahedral Finite-Volume Solutions to the Navier-Stokes Equations on Complex Configurations, Neal T Frink

this is the link:

http://ntrs.nasa.gov/search.jsp?R=24...57c1%26N%3D123

Check the formula at (3) this is the more robust.

Paul

gonski · April 1, 2009, 02:26

Hi Paul,
Thank you very much!
Is it possible for me to use weighting method to interpolate velocity from a cell as well as 6 neighbors (if hexadra is considered ) to some point in the cell? Or I need to do the weighting interpolation twice to achieve this. will you please present some comments here?
Cheers,
Gonski

RenardP · April 1, 2009, 08:19

I'm not sure I understand correctly your question. Are your using a dual mesh or some form of staggered grid for solving your problem ?

gonski · April 1, 2009, 08:46

Sorry! I did not explain my problem clearly enough.
Now I have two things to be done as follows:
(1) To get velocity on grid nodes from cell center which have all flow information in my code. This is mainly for visualization
(2) To get local velocity on particle position for calculating particle trajectory. Here the particle can reside anywhere in a cell.

As you suggested, I can achieve the first item by weighting interpolation.
To calculate local velocity, I may have two choices when I know which cell the particle of interest is residing:
(a) I can interpolate node based data (calculated in task ( 1)) to particle position. This means that I do two weighting interpolations. Here, I need to only consider velocities on 6 nodes of the host cell containing the considered particle.
(b) I can directly interpolate cell-based data to particle position. This means that I only do weighting interpolation once. Here, I need to consider velocity on the host cell and its 6 neighbors will be used.

Which is better to get local velocity to calculate particle trajectory?

RenardP · April 1, 2009, 14:12

OK,

I have experience only on point 1) - for this I can tell you that the weight based interpolation will work.

For point 2) I don't know what will be the best method, I suggest you to create an "academic example" - give some analytical form to your velocity field (I suppose your particle has the same velocity as the surrounding fluid), and test this on some simple meshes (up to 6-10 points), then check if the weight based formula gives you an accurate results on this mesh.

I think it will work, but you better check it.

Paul

gonski · April 1, 2009, 22:40

Hi paul,
Than you very much for your kind helps. Other guys who give me helps are also greatly appreciated. From the simulations outputs I have got in different systems. Now my code works well with inverse weight interpolation. As the simulations would finish in three weeks at least due to high solid loading, I cannot tell this interpolation can improve my code. Whatever, simulations are running as expected. This already means a lot.
Thanks again.
All the best,
Gonski

March 31, 2009, 18:37	cell-based data to node-based data	#1
gonski New Member david.notay Join Date: Mar 2009 Posts: 20 Rep Power: 17	Hi all, My CFD code provides fluid velocity on cell center. But I need velocity on grid nodes for some special calculation. So, which scheme is the best for me to interpolate the cell-based data to grid node? Your hints would be greatly appreciated. Cheers, Gonski

Similar Threads
Thread	Thread Starter	Forum	Replies	Last Post
edge/face based data structure unstructured grid	sundeep	Main CFD Forum	0	April 25, 2007 12:36
which is better? vertex based or cell centered?	neil	Main CFD Forum	1	April 23, 2007 16:54
interpolation method on unstructured meshes	harry	Main CFD Forum	9	October 3, 2006 01:10
About node based calculation	Lohen	FLUENT	2	January 28, 2004 10:50
Cell Vertex Based FVM	A. S.	Main CFD Forum	1	January 28, 2002 07:46

March 31, 2009, 20:49		#2
gory Member Join Date: Mar 2009 Posts: 32 Rep Power: 17	Volume-weighted average, Least-squares fit (possibly with weights), etc. Or generate a dual grid whose nodes coincide to the cell centers of the original grid?

March 31, 2009, 21:08		#3
gonski New Member david.notay Join Date: Mar 2009 Posts: 20 Rep Power: 17	Now I try to use gradient interpolation, i.e. interpolate velocity from center of cells around the considered node to the node, then average it. the gradient is calculated by a leas-square or guass method. I guess if gradient is based on least-square method, it should be least-square fit mentioned by you? I found this method produce strange results on boundaries.

March 31, 2009, 21:58		#4
RenardP Member Join Date: Mar 2009 Posts: 62 Rep Power: 17	You can search on NASA reports site after Neal T Frink, he presents a simple method for interpolating the cell centered values to the grid nodes. His method works very well on unstructured grids.

April 1, 2009, 00:10		#6
RenardP Member Join Date: Mar 2009 Posts: 62 Rep Power: 17	Tetrahedral Finite-Volume Solutions to the Navier-Stokes Equations on Complex Configurations, Neal T Frink this is the link: http://ntrs.nasa.gov/search.jsp?R=24...57c1%26N%3D123 Check the formula at (3) this is the more robust. Paul

April 1, 2009, 02:26		#7
gonski New Member david.notay Join Date: Mar 2009 Posts: 20 Rep Power: 17	Hi Paul, Thank you very much! Is it possible for me to use weighting method to interpolate velocity from a cell as well as 6 neighbors (if hexadra is considered ) to some point in the cell? Or I need to do the weighting interpolation twice to achieve this. will you please present some comments here? Cheers, Gonski

April 1, 2009, 08:19		#8
RenardP Member Join Date: Mar 2009 Posts: 62 Rep Power: 17	I'm not sure I understand correctly your question. Are your using a dual mesh or some form of staggered grid for solving your problem ?

April 1, 2009, 08:46		#9
gonski New Member david.notay Join Date: Mar 2009 Posts: 20 Rep Power: 17	Sorry! I did not explain my problem clearly enough. Now I have two things to be done as follows: (1) To get velocity on grid nodes from cell center which have all flow information in my code. This is mainly for visualization (2) To get local velocity on particle position for calculating particle trajectory. Here the particle can reside anywhere in a cell. As you suggested, I can achieve the first item by weighting interpolation. To calculate local velocity, I may have two choices when I know which cell the particle of interest is residing: (a) I can interpolate node based data (calculated in task ( 1)) to particle position. This means that I do two weighting interpolations. Here, I need to only consider velocities on 6 nodes of the host cell containing the considered particle. (b) I can directly interpolate cell-based data to particle position. This means that I only do weighting interpolation once. Here, I need to consider velocity on the host cell and its 6 neighbors will be used. Which is better to get local velocity to calculate particle trajectory?

April 1, 2009, 14:12		#10
RenardP Member Join Date: Mar 2009 Posts: 62 Rep Power: 17	OK, I have experience only on point 1) - for this I can tell you that the weight based interpolation will work. For point 2) I don't know what will be the best method, I suggest you to create an "academic example" - give some analytical form to your velocity field (I suppose your particle has the same velocity as the surrounding fluid), and test this on some simple meshes (up to 6-10 points), then check if the weight based formula gives you an accurate results on this mesh. I think it will work, but you better check it. Paul

April 1, 2009, 22:40		#11
gonski New Member david.notay Join Date: Mar 2009 Posts: 20 Rep Power: 17	Hi paul, Than you very much for your kind helps. Other guys who give me helps are also greatly appreciated. From the simulations outputs I have got in different systems. Now my code works well with inverse weight interpolation. As the simulations would finish in three weeks at least due to high solid loading, I cannot tell this interpolation can improve my code. Whatever, simulations are running as expected. This already means a lot. Thanks again. All the best, Gonski