I have set up my grid in pre-BFC and attempted to do the modelling in Fluent. After inputting the necessary parameters, I requested the no. of iterations and the following message appeared "Rapid Divergence detected *aborted* " What are the possible reasons for this happening ? The geometry is simple and the fluid is water.

(1). Nowadays, these large complex programs are designed to handle very complicated geometry, so, I think, it does not like your simple case and plain water. You have to upgrade it to at least German beer or French wine level before it will start responding to your challenge. (2). Well, it is user-friendly, and the message simply says " you are wasting your time, come back next century." (3). From your case, I can say that fooling around with a black box is not always fun. (4). These programs usually don't like strangers. So, you can first pretend that you are a friend of the program, by running a sample case first. If the program accepts the sample case, then you are O.K. (5). Now the next step is to make a very small change to the sample case, but be careful not to let the program know about it. As I said, the program does not like strangers, so, you have to keep learning its language. (6). My suggestion is " look for a working sample case first, then make a secret change to the sample case ,without provoking the program's self-defense shield". (7). The best way to " handle a program " is to show the program your strength, that is you are independent and can live without the program. (8). Maybe, the program has AI capability and it is studying your behavior right now. (9). So, you got rejected by a program?..... We are not in 2001 yet. Relax....... and make your next move.

Maybe you setup the geometry in mm and forgot to do the right scaling to meters

U might try and set the underrelaxation parameters and also look at initializing the solution.....

Could be down to grid quality, as usual. Be careful not to specify too large or too small a near wall cell size on walls, if you're modelling turbulent flow. Make sure Y+ is with 30-500 and not less than 11. I've found CFX is very sensitive to this, when using the K-e turbulence model, but i'm not too sure about Fluent.Good luck

James

(1). If you are still having divergence problem, I think, you need some help. (2). Try to run a 2-D channel flow problem first. For this problem, all you need is four points to define the rectangular domain. You should be able to do this in preBFC/2-D mode easily by setting 4 points, and create straight lines between two points for the inlet, outlet, upper wall and bottom wall. You can then distribute nodes on these lines using uniform distribution or non-uniform distribution. (3). Anyway, at this point you should be able to generate the mesh. After that, just follow the normal procedure to set and select the options and conditions using the menu items from left to right.(4). So, try out this simple 2-D channel flow first. If it still does not work, you must follow the sample cases first to make sure that you can get the sample cases running. (5). I started using Fluent in early 96. I used the program almost everyday, for over two years. I did not attend the training sessions, but I followed the sample cases one by one, on my own. And I was able to obtain very complicated 3-D flow field through preBFC, tgrid, UNS and Rampant. I did not run into any convergence problem. ( convergence for 3-D Rampant solution was very slow though.) The main problem I had was the 3-D automatic volume mesh generation using tgrid. So, most of the time, I was on the phone talking to the support engineer on the mesh issues. (6). With Gambit, it should be much easier for users to generate a successful 3-D mesh. (7). Fluent is one of the easiest and most user-friendly code I have ever used.

Dear FLUENT user,

It's not clear to me which version of FLUENT you're using. In many cases, checking problem set-up, including boundary conditions, etc. would help. If your mesh is highly skewed, you may want to use less aggressive under-relaxation factors.

If all this does't help, please contact your support engineer.

If your problem is set up and initialized correctly, try to switch the convergence check off for the first few iterations!