CFD Online Discussion Forums

CFD Online Discussion Forums (http://www.cfd-online.com/Forums/)
-   FLUENT (http://www.cfd-online.com/Forums/fluent/)
-   -   Accessing neighbouring cells in UDF (http://www.cfd-online.com/Forums/fluent/29526-accessing-neighbouring-cells-udf.html)

Hiranya Nath March 19, 2002 01:36

Accessing neighbouring cells in UDF
 
Hi!

I have to write an UDF for momentum source that will replace a hole - the problem is to identify the cells those will lie in the hole. The user should provide two data for this: 1. x,y,z location of the hole and, 2. diameter of the hole. Using these info how can I identify the cells that will represent the hole. I request you kindly a solution for this - I will be very much grateful to you.

Thanks, Hiranya

Greg Perkins March 21, 2002 20:18

Re: Accessing neighbouring cells in UDF
 
Hi,

you can use the C-CENTROID function to determine whether the current cell is located inside or outside of the hole. Use this as:

real x[ND_ND];

C_CENTROID(x,cell,thread); x0 = x[0]; y0 = x[1]; z0 = x[2];

Of course, this is approximate since part of the cell may be outside the perimeter of the hole even though the centroid is within it.

Greg

Hiranya Nath March 22, 2002 05:28

Re: Accessing neighbouring cells in UDF
 
Hi,

Thanks for your suggestion. There is a problem - I need to restrict the cell searching in the plane of the hole; otherwise the searching would be within a sphere of the hole-diameter which, I do not want. How can I seacrh the cells in the plane of the hole. Please give some advice. May be for this the user has to give some more inputs.

Thanks in advance, Hiranya


Greg Perkins March 22, 2002 06:01

Re: Accessing neighbouring cells in UDF
 
Not sure exactly what you mean.

Do you mean - 1. you want to only loop over cells in a certain x-y plane?? or 2. you want to ensure you only pick cells within the hole independent of say the z-co-ordinate value??

For 1. there is no easy answer - unless you specifically setup your mesh with a cell zone around the location of the hole. Otherwise you will need to loop over the whole domain (actually there are some smart tricks one could use...eg. do a calculation initially and store the cells in the local vicinity of the whole in your own list)

For 2. this is relatively easy, since you just ignore the z value returned by the C_CENTROID function and make comparisons in only the x-y plane.

Greg


Hiranya March 22, 2002 07:58

Re: Accessing neighbouring cells in UDF
 
Sorry for the lack of clarity in my mail.

Actually the plane, on which the hole lies, is an arbitrary plane. It may coincide with say x-y plane or it may be at some angle(s) with the principal cartesian planes. So I want to basically traverse along the plane of the hole (instead of say x-y plane). So I think I would need some some kind of direction vector along the plane. What do you say?

I found the trick that you mentioned quite interesting; I have got some idea about it but it is not fully clear how can I execute DEFINE_SOURCE macro only for my own list of cells. Would you please give me some more insight on this.

Thanks, Hiranya

Greg Perkins March 23, 2002 00:29

Re: Accessing neighbouring cells in UDF
 
1. you may need to consider what you're going to do with the results etc. to determine the optimum method. ...if you are going to use a source term then you can have two options: a) calculate the right value in the cell when the source is called, or b) combine with a DEFINE_ADJUST which calculates the source for every cell and then saves to C_UDMI location. In DEFINE_COURSE you simply return the pre-computed value.

Doing the first, implies looping over all cells. If you do the later, I suspect you will still find it easiest to loop over all cells and perform a calculation.

Restricting the looping over cells to an arbitrary plane is possible. I would imagine, you need to loop over the faces/cell-neighbours of the current cell and apply your direction vector to see which nieghbouring cells to go to next, as you mentioned. Its likely to be a little more tricky....

If I were you, I'd see if I could to it simply with a simple algorithm before doing any optimisation. It might run fast enough to ignore coding any special tricks...

Greg


All times are GMT -4. The time now is 21:56.