grid dependence, flow over a cylinder (inside a wind tunnel)
 

greetings,

this is my first thread, so im kind of nervous :) especially because im new to CFD, ok here it goes:

viscous, steady state ,flow over a cylinder, subsonic, (grid generated from icem cfd exported to cfx)

when you do a grid independence study ( from what i read) you need to increase the number of nodes on your grid until your answers dont change, what i dont understand is, wich variable should i choose (velocity presure) and what value of the variable ( max? avarege?), plus my teacher ask us to do a plot of numver of nodes (x axis) vs any variable(y) in order to see the grid dependence clearly, is there a way to do this directly from cfx ? if not how can i export values of a variable to a table on excel or TXT.

thanx in advance for the fast reply, and i apologyze in advance for mi english.

Hi
the variables that are important for your project (directly or indirectly) and/or you want to study them.
yes
use chart in cfx-post and export data, you can edit and use it in a program like excel

please search cfx help in the first step
good luck

For flow over a cylinder a typical thing to go for mesh convergence on would be the drag force.

thanx, that was fast, i know that my question is noobish, thanx again, im gonna try with the chart, its just that i though that i could save the variable values (velocity) directly to the table option on cfx, o an Glen im gonna take your advice and put the drag force im my work :)

you can do it

ups i made a mistake, how can i delete a post, i see that edit has edit/delete but i don see it ( i posted something that somebedy else help me earlier )

as i know the supporters disabled this ability
how is finishing the post with the best answer that you have found

great idea
 

so the best answer was:

if you wanna find grid independence you improve the quality of the mesh, how? well you increment the number of nodes depending on your criteria, for example my teacher gave me this criteria

Ni> 1.5(Ni-1)

Ni= number of nodes
Ni-1= number of nodes prior to the improvement

so your adding 50% more nodes than before, or you could just double and double and keep doubling nodes until (in my case)velocity of the fluid reachs the error criteria [ ((Ri-(Ri-1))/100) *100 ] mine was 2% of error.

Ri= present result (example : max velocity magnitude)
Ri-1=results before you increment the number of nodes

o an you can export the values of your variable by creating streamlines and then creating a char of your variable (example: velocity vs X)
that way you can see the behavior of velocitys x component and then export from chart the values of that component into any excel page.

thanks again for the help you will be hearing alot of me :):)

In many cases the increment of mesh size you suggest is not large enough to pick up a difference between mesh sizes. For instance, if the elements are cubes, then the next element size is a ratio of 1.15 less than the first. In my experience this is not a large enough difference.

When I do mesh refinement studies I double the mesh element lengths. This means from a unit cell, I do a cell of 0.5 edge length and 2.0 edge length. This means my number of elements goes up and down by a factor of 8.

This reference gives some good detail about what's going on:
http://journaltool.asme.org/Template...umAccuracy.pdf

And for the detailed theory see "Computational Fluid Dynamics" by Roache.

Oh, and as a moderator I can delete posts. But I don't do that unless it is offensive, advertising or not appropriate for some reason.

you are absolutely right, sorry for making such a bold statement, that reference is real good, now i can know for sure if what im doing is right or wrong ( you shoul post that reference in the general forum, its really good information) .

i forgot to write that the procedure i wrote was for a unstructured grid ( basic design) and its just an aproach for an unexperienced (like myself) CFD user :)

Pd: im gonna blow my teachers mind with that guide :)