CFD Online Discussion Forums

CFD Online Discussion Forums (http://www.cfd-online.com/Forums/)
-   Main CFD Forum (http://www.cfd-online.com/Forums/main/)
-   -   grid independence (http://www.cfd-online.com/Forums/main/15377-grid-independence.html)

anne July 2, 2008 08:40

grid independence
 
I am checking grid independence in my case, which is a external air flow on the airplane. i am just wondering how much difference is acceptable among different mesh sizes. because no matter how fine the mesh is, the solution can't be exactly the same. there are always a tiny difference. it there any rule?

if any of your guys would like to have a look of my solution, i will send it to you.

many thanks

mettler July 2, 2008 09:44

Re: grid independence
 
I believe there are a few ways to do grid checks. I did mine by comparing an average value of 'planes' that I had set up within my grid with different grids. For example, I would check the average temperature of the plane at different grid sizes. When I plotted the value vs number of cells I could see when the grid was too coarse. I would also check the value vs the changed grid count. For example, I would compare the value obtained at grid A with 1000 cells to that of grid b with 500 cells and have a criteria set that it would have to meet - ie..can only change by .001. Grid check used to be used to help with computational time as well as accuracy, but since computers have gotten so much more powerful it is really just an accuracy check now. that is my 2 cents worth anyway

anne July 3, 2008 04:57

Re: grid independence
 
thanks a lot, mettler,

your idea is really helpful for me. I am just wondering another thing, the criteria, is it a absolute one or a relative one. I guess it should be a relative one. what's your definition?

thanks again

mettler July 3, 2008 07:02

Re: grid independence
 
my criteria depended on what I was looking at. If I were looking at temperature my results would be something like this.

Cell Count Temperature Plane ---------- ----------------- 5000 300.514 4000 300.515 3000 300.514 2000 300.485

Then I would look between 3000 and 2000 if I was trying to optimize it.

Again, this was just done to help with computing power. From the numbers above 4000 would be just fine, since it didn't really change above that or below it. In the case of temperature anyway.

anne July 3, 2008 07:41

Re: grid independence
 
ok, I see.

I also wonder if you have tried to compare the variables difference with your RMS residual. Do we have to keep the variables difference in the order of residual, say 1e-5 or 1e-6.

many thanks, really

anne

mettler July 3, 2008 08:30

Re: grid independence
 
I think that will be up to you. I wouldn't waste too much time on this though. I would show that your answers do not change appreciably between different numbers of cells and then just go from there. If I remember correctly, I compared my variables to the order of my convergence criteria. You just need to find the cell count in which your results start to change and then increase that number of cells. The number of cells is going to be dependent on what you are looking at too. If it is a complicated geometry, then you are going to need a lot of cells to capture everything that is going on.

anne July 3, 2008 11:06

Re: grid independence
 
thanks a lot, mettler.

I will follow your guide and see what will happen.

all the best

anne

mettler July 3, 2008 11:13

Re: grid independence
 
a plot is the best way to see it, and to show it (ie to a professor or something like that). pick a variable and record its value with different cell/grid counts and plot. You will get a horizontal line until the grid gets too coarse and then it will fall or go up - meaning the results are bad. You want your cell count to be on that straight line. My X-axis was my cell number and the Y-axis is the variable result. It gives you a nice graph to show how you arrived at the cell count that you did

littlelz July 3, 2008 16:59

Re: grid independence
 
mettler,

thanks again.

that's exactly what I am doing, like you said, the straight line will go up or go down, and it's a very nice show.

best regards,

anne

Mel August 6, 2008 09:38

how to model a entrifugal pump using gambit
 
hello..

i am working on modelling a centrifugal pump (the impeller and its spiral casing).can anyone help me how to do this?i really need steps or procedures to draw the pump impeller in gambit.

thanks.

mettler August 6, 2008 09:56

Re: how to model a entrifugal pump using gambit
 
don't draw it in Gambit..draw it in SolidWorks or ProE and import it.

Ser tang June 24, 2010 02:44

Quote:

Originally Posted by anne
;55700
I am checking grid independence in my case, which is a external air flow on the airplane. i am just wondering how much difference is acceptable among different mesh sizes. because no matter how fine the mesh is, the solution can't be exactly the same. there are always a tiny difference. it there any rule?

if any of your guys would like to have a look of my solution, i will send it to you.

many thanks

Dear Sir ,

Can you send your example to me as I have no ideas how to do grid independency test ? Thanks you very much . Your cooperation is highly appreciate. :D

Ser tang June 24, 2010 02:48

DeaR Anne ,

For your information , I would like to view , how you do your grid independency test. My e-mail is ser.tang@yahoo.com , hope to receive the feed back from you soon. Thanks you very much . :D

Ser tang June 24, 2010 03:17

Quote:

Originally Posted by mettler
;55707
I believe there are a few ways to do grid checks. I did mine by comparing an average value of 'planes' that I had set up within my grid with different grids. For example, I would check the average temperature of the plane at different grid sizes. When I plotted the value vs number of cells I could see when the grid was too coarse. I would also check the value vs the changed grid count. For example, I would compare the value obtained at grid A with 1000 cells to that of grid b with 500 cells and have a criteria set that it would have to meet - ie..can only change by .001. Grid check used to be used to help with computational time as well as accuracy, but since computers have gotten so much more powerful it is really just an accuracy check now. that is my 2 cents worth anyway

Hi mettler,

I would like to know how to do grid independency test? your ooperation is appreciated . Hope to hear from you soon. Thanks you very much.

Ser tang June 24, 2010 09:02

Quote:

Originally Posted by Ser tang (Post 264290)
DeaR Anne ,

For your information , I would like to view , how you do your grid independency test. My e-mail is ser.tang@yahoo.com , hope to receive the feed back from you soon. Thanks you very much . :D

Dear Sir,

can you guide me how to do independency test ? thanks alot :D :) ;)

tanmoy95 June 24, 2010 09:07

reply:
 
hi !..to make a grid independancy test make the mesh finer ..increasing the no of meshes 2^n times the previous, and analyse everything..if u see the data does not change much,( a predefined value say 0.05) then u say the results have decayed to convergence in response to grid independancy test!..thank u :)

Ser tang June 24, 2010 09:08

Quote:

Originally Posted by mettler
;55707
I believe there are a few ways to do grid checks. I did mine by comparing an average value of 'planes' that I had set up within my grid with different grids. For example, I would check the average temperature of the plane at different grid sizes. When I plotted the value vs number of cells I could see when the grid was too coarse. I would also check the value vs the changed grid count. For example, I would compare the value obtained at grid A with 1000 cells to that of grid b with 500 cells and have a criteria set that it would have to meet - ie..can only change by .001. Grid check used to be used to help with computational time as well as accuracy, but since computers have gotten so much more powerful it is really just an accuracy check now. that is my 2 cents worth anyway

Dear Sir,

Can you guide me how to do the grid independency test as I don't know how to do it. Is it , I need to obtain the " grids " , meaning using the number of small grids devide with the bigger number of grids and the multiply by 100% , so I can obtain the results the independency test results . Is this consider as the independency test? Your cooperation is highly appreciate . :) :D ;) :* :P :( :O :/

Ser tang June 24, 2010 09:14

Quote:

Originally Posted by tanmoy95 (Post 264346)
hi !..to make a grid independancy test make the mesh finer ..increasing the no of meshes 2^n times the previous, and analyse everything..if u see the data does not change much,( a predefined value say 0.05) then u say the results have decayed to convergence in response to grid independancy test!..thank u :)

Dear Sir,

I still not sure which icon i need to click ,so I can do the independency test.....meaning of this statement?? " a predefined value say 0.05) then u say the results have decayed to convergence in response to grid independancy test" .I hope you can guide me deeply as , my foundation is quite weak. Your cooperation is highly appreciate . Thanks you very much. :* :) :D :P :( :/

Ser tang June 24, 2010 09:21

Quote:

Originally Posted by tanmoy95 (Post 264346)
hi !..to make a grid independancy test make the mesh finer ..increasing the no of meshes 2^n times the previous, and analyse everything..if u see the data does not change much,( a predefined value say 0.05) then u say the results have decayed to convergence in response to grid independancy test!..thank u :)

Dear Sir ,

Where can I see predefine value state in fluent software ?? :(

tanmoy95 June 24, 2010 09:50

reply:
 
i just talked about the theory ..crudely speaking ,to make a grid indpendance test what u need to do is increase the mesh quantity manyfold, and see if u find any considerable change in the data..if not then ur solution has reached grid independance !..

i dont know if there is a switch to make the test..u can try it out by remeshing ur model every time in gambit !...:-):)


All times are GMT -4. The time now is 16:45.