CFD Online Logo CFD Online URL
www.cfd-online.com
[Sponsors]
Home > Forums > FLUENT

Convergence in 3D simulation

Register Blogs Members List Search Today's Posts Mark Forums Read

Reply
 
LinkBack Thread Tools Display Modes
Old   October 19, 2000, 07:40
Default Convergence in 3D simulation
  #1
jarbas a. da rosa
Guest
 
Posts: n/a
I'm using the FLUENT 5 to simulate turbulent airflow with mixed convection in a three-dimensional room, in my case the heat sources are two persons, two computers and six fluorescent lamps. I'm not reaching the convergence for the steady-state solution (this solution exist according to the paper used to comparison), the residuals begin to oscilate and stop to decrease. I already tried change the values of underrelaxation and use a time-dependent approach, but the problem continues. Could anybody please help me?

Jarbas
  Reply With Quote

Old   October 19, 2000, 08:25
Default Re: Convergence in 3D simulation
  #2
John C. Chien
Guest
 
Posts: n/a
(1). I must say that if you need reliable results you have to run the test. This is the only approach, and in my case, we are also going to run test because commercial cfd solutions are simply not reliable nor accurate at the present time. (2). This is consistent with my previous messages. It is nothing new. After having said that, it is possible to get some information from the cfd results, if you are highly trained in your field. But the interpretation of the cfd results is very tricky, and requires a lot of luck. (3). This is the reason why I made the prediction sometimes ago that a company using only commercial cfd codes will eventually die. (in many cases, the company is already nearly dead.) Many years ago, a friend of mine showed me that his simulation of the flow in a room would change directions if he looked at it next day on the worstation. (4). If you think cfd is just like finite element analysis (structure), you are just like one of our project manager, he was thinking that with the CAD geometry, the cfd solution could be done right away. (this is the reason why his project will fail next year, when these cfd results finally are used in the design. actually, in many cases, only a rough guess were used, not even the cfd results,because the number keeps changing as a function of mesh.) (5). My suggestion is: write your own code, and if you still have problem, do something else. Even in my case, the answer is to run the test if the solution is important.
  Reply With Quote

Old   October 19, 2000, 12:19
Default Re: Convergence in 3D simulation
  #3
Evan Rosenbaum
Guest
 
Posts: n/a
You need to determine if you have a properly posed problem that won't converge or an improperly posed problem that can't converge. Don't automatically look to the solver first.

In your problem, for example, buoyancy effects will be significant. Check your applied gravity and your density vs. temperature data. Do not use the Boussinesq approximation for problems with non-negligible density gradients. This is just an example of the things to check, before turning to the solution controls.

If everything look right with your inupts, converge the solution as far as possible. Do the velocity contours change a lot, or oscillate, every few iterations? This could indicate that the global solution is being overwhelmed by the local effects. Or does the solution appear to be approaching one result? Maybe the mesh isn't fine enough.

I hope this helps you somehow. Problems with strong buoyancy forces, like your probelm, require extra attention to the interaction of global and local effects.
  Reply With Quote

Old   October 20, 2000, 08:00
Default Re: Convergence in 3D simulation
  #4
Sugen Chetty
Guest
 
Posts: n/a
Well, I did have that problem with my 3D simulations. A dynamic solution took in the order of a day for a time period, but some errors were still occuring. (mainly pressure) This could be due to some error on part as Evan has mentioned. However, I have accomplished my simulation. Problem is I can't exactly remember how. I know that increasing the minimum residual sum does reduce solution times. Probably can tighten this once it starts converging. I did play around with the relaxation parameters and the number of sweeps. Should read up on ways of solving problem in manual. Thats your best bet. Then try each way and note down what happens. That way you can repeat your method when you do solve your problem.

Cheers
  Reply With Quote

Old   October 23, 2000, 10:01
Default Re: Convergence in 3D simulation
  #5
jarbas a. da rosa
Guest
 
Posts: n/a
I just try every suggestions of the manual, but the problem continue.

Jarbas
  Reply With Quote

Old   October 23, 2000, 11:33
Default Re: Convergence in 3D simulation
  #6
John C. Chien
Guest
 
Posts: n/a
(1). If you are still interested in solving your problem, you need to do it step-by-step. (2). You can do this with one object in the room, and try to obtain the converged solution. Once you have done this, then you can add one more object in the room. And continue the exercise. (3). So, don't try the complex problem right away. You need to find out whether the code can solve the simple problem you are interested in. If you still have problem with one object in the room, try to replace it by a simple block to make it easier to generate the mesh. (I don't know what's wrong with your problem or the code, but with this approach, you should know where you stand.)
  Reply With Quote

Reply

Thread Tools
Display Modes

Posting Rules
You may not post new threads
You may not post replies
You may not post attachments
You may not edit your posts

BB code is On
Smilies are On
[IMG] code is On
HTML code is Off
Trackbacks are On
Pingbacks are On
Refbacks are On


Similar Threads
Thread Thread Starter Forum Replies Last Post
Cyclone Simulation Convergence Problem Sal FLUENT 8 December 17, 2014 08:46
problem with Min/max rho tH3f0rC3 OpenFOAM 7 February 23, 2013 06:37
Convergence of CFX field in FSI analysis nasdak CFX 2 June 29, 2009 01:17
How stop simulation if convergence is not reached? Andreas FLUENT 0 October 16, 2007 12:35
convergence for an Unsteady simulation Palani Velladurai FLUENT 1 March 19, 2007 11:56


All times are GMT -4. The time now is 08:44.