
[Sponsors] 
May 2, 2006, 15:24 
Solution control  discretization

#1 
Guest
Posts: n/a

Hi experts! Does anybody know the different calc. schemes (First order upwind, Second order upwind, Power law, Quick and Third order MUSCL)available in Discritization for momentum? How each of them effect yr result? I am working with VOF, droplet hitting flat surface, which one is the best scheme for me? Thx.


May 3, 2006, 10:04 
Re: Solution control  discretization

#2 
Guest
Posts: n/a

Hi Aireen, 1st order is subjected to a numerical diffusion thus the result may not accurate especially when you are working with tet grid. 2nd order is quite accurate but your simulation will be less stable, anyway it is a wise discretisation choice considering it accuracy. MUSCL and Quick is a 3rd order accurate scheme, but be sure there is no flux limiter in MUSCL so you simulation residual might diverge. Quick in another hand is only applicable to hex grid, in tet grid it will only 2nd order accurate. I didn't find any advantage of using quick in hybrid grid. I never try power law so I cannot comment on that. I've do VOF before in 1st order and 2nd order scheme in a very good hex mesh. The result is comparable with the experimental data. For your problem it might good to solve it using 2nd order discretisation. Hope it help.


May 3, 2006, 10:59 
Re: Solution control  discretization

#3 
Guest
Posts: n/a

Hi J.Gimbun, Thx for yr info. I think I'm going to try using 2nd order discretization. I have another question, how does this 2nd order discretization compare to the 2nd order implicit, found in the SOLVER panel? Again, thx a bunch!


May 3, 2006, 12:41 
Re: Solution control  discretization

#4 
Guest
Posts: n/a

I use the 2nd order implicit in solver panel as well in my current study because some people from Fluent (Andre Bakker) using that. To be honest I don't really understand about it but I guess it might be applied within the 'entire' time step iteration. good luck.


May 3, 2006, 22:23 
Re: Solution control  discretization

#5 
Guest
Posts: n/a

I've just asked one of the cfd professor who is familiar with FLUENT, he said 2nd order implicit is always recommended for a better accuracy. BTW, did you do any study to find an optimum number of meshes? I'm doing it right now (since my secondary advisor recommended me to do so), but it's very time consuming. Thx, good luck to you too.


May 4, 2006, 03:38 
Re: Solution control  discretization

#6 
Guest
Posts: n/a

hi,Gimbun I am wondering how you can use the 2nd order implicit in solver panel with VOF method. I try to select it then click "ok". bur when I check it again, I find it is still 1st order implicit. the manual of fluent said vof will use the 1st order implicit scheme for time discretization.
Anything wrong with me? Thanks! 

May 4, 2006, 04:17 
Re: Solution control  discretization

#7 
Guest
Posts: n/a

Maybe I'm wrong, ya there is some bug in Fluent in VOF, for example if you set your BC as a phase 1 then you will find all other BC also phase 1! Similar problem with species modelling when I performed my gas phase reaction modelling 2 year ago. What I mean is when we click the 2nd order implicit & iterate then after that we found the 1st order is thicked instead. To be honest in such situation I'm not very sure which scheme is employed by Fluent for calculation, I believe that this is a bug that Fluent have to solve. Anybody have an idea about that. Cheers,


May 4, 2006, 04:25 
Re: Solution control  discretization

#8 
Guest
Posts: n/a

hi, do you know whta is the interface reconstruction algorithm use in Fluent vof method with the "Euler explicit"scheme? I use the Georesconstruction scheme first, but after some time steps ,it is unstable. The "Euler explicit" can be stable but I am not sure the interface reconstruction method it used. I am a fresher to VOF. Can you share your experience with me? Thanks!


May 4, 2006, 04:40 
Re: Solution control  discretization

#9 
Guest
Posts: n/a

Ya optimum mesh study or grid independent study must be carried out especially if you doesn't have the experimental data to do validation. I also perform a grid independent study in my case although I have a 3D LDA and PIV data to validate them. I start from a grid size of 150000 to 600000 cells (halfdomain of my study). I found that u, v & w will be predicted accurately at considerably coarse grid but not a turbulent kinetic energy. I didn't find the grid independent study is time consuming, anyway it depend on the PC/workstation and grid size of which you work with.


May 4, 2006, 04:49 
Re: Solution control  discretization

#10 
Guest
Posts: n/a

To be honest I'm not good in VOF too, I just helping my friend to solve one of his PhD problem and accidently succesful to do that. The problem is a fairly simple, is just about to find the right pressure to 'push' the water out from filter pore and we manage to get a 'reasonable agreement' with the experimental observation. There are ample of reading material about the VOF theory in Fluent and any CFD book that might help you. It is also worth to try each method and compare them with experimental data. regards, Jolius


May 5, 2006, 11:05 
Re: Solution control  discretization

#11 
Guest
Posts: n/a

yes, i have the same problem, it just can't take the 2nd order implicit.


May 5, 2006, 11:11 
Re: Solution control  discretization

#12 
Guest
Posts: n/a

With that number of cells, what the time step size did you use? How big is your actual/physical domain? My actual domain is 60x120 micron, with 90x180 meshes, it took me a few days to simulate 1s event. I'm using 1e7 time step size. The time is really killing me!


May 5, 2006, 16:55 
Re: Solution control  discretization

#13 
Guest
Posts: n/a

Hi Aireen, your domain just have about 16k cells. I suppose your problem should be solved within a minute if is it 2D. There must be a reason to choose a time step i.e. in moving mesh time step should be in correspondance to 1 cell size movement. I didn't see the need of 0.1us time step in your case. I have try a time step of 0.0001s to almost similar domain with yours (100micron x 100micron) before and still I can see a reasonably good result. My suggestion is just try a slightly bigger time step first and reduce them if not appropriate.


May 5, 2006, 17:45 
Re: Solution control  discretization

#14 
Guest
Posts: n/a

The reason why I used smaller time step size is because the solution did not converge. I gave an error mesage saying something like floating point error. Do you know how to solve this problem without reducing the time step size?


May 6, 2006, 07:48 
Re: Solution control  discretization

#15 
Guest
Posts: n/a

in unsteady simulation you don't have to border about convergence, just make sure your residual is low enough i.e. below 1e4 every iteration then it should be fine...


May 6, 2006, 11:05 
Re: Solution control  discretization

#16 
Guest
Posts: n/a

Do you mean bigger that 1e4 such as 1e3 or lower than 1e4 such as 1e5. What is the appropriate number of iteration? Please help me. If it is true (doesn't need to be converged), it could save my 'life'. Thx.


May 7, 2006, 06:38 
Re: Solution control  discretization

#17 
Guest
Posts: n/a

In unsteady simulation you only need to make sure that your residual is lower than 1e3 (minimum convergence) at each time step. With that your solution can be assumed fairly converged. Hope it help.


Thread Tools  
Display Modes  


Similar Threads  
Thread  Thread Starter  Forum  Replies  Last Post 
grid dependancy  gueynard a.  Main CFD Forum  19  June 27, 2014 21:22 
Redefining variables in fluent instead using limits (control solution)  akhlaghi  FLUENT  0  February 19, 2011 07:14 
Exact solution of Burgers equation  mcaro  Main CFD Forum  3  January 25, 2011 07:46 
Discretization problem in Solution controls  Tamm  FLUENT  0  July 15, 2005 07:07 
Wall functions  Abhijit Tilak  Main CFD Forum  6  February 5, 1999 02:16 