CFD Online Discussion Forums

CFD Online Discussion Forums (https://www.cfd-online.com/Forums/)
-   FLUENT (https://www.cfd-online.com/Forums/fluent/)
-   -   A STRANGE PROBLEM!!! (https://www.cfd-online.com/Forums/fluent/31119-strange-problem.html)

Fanadam March 19, 2003 06:59

A STRANGE PROBLEM!!!
 
Hi, everybody! I'm facing a strange problem! If I set the initial velocity of the jet at the inlet is 600m/s, everything is OK. But if I change the value to 900m/s, the system will stop working and prompt me " Deivergence detected in AMG solver, X-momentum". I don't know how to solve it, and 900m/s is necessary.

Who can help me? Thanks a million!!!

kaiser March 19, 2003 07:12

Re: A STRANGE PROBLEM!!!
 
this problem is very common, ordinary, and traditional! lol!To solve it, change some parameters, even the solution control.

mateus March 19, 2003 08:51

Re: A STRANGE PROBLEM!!!
 
You can also try to increase velocity in smaller steps untill you reach 900 m/s...

Ming March 19, 2003 10:34

Re: A STRANGE PROBLEM!!!
 
You can decrease relaxation factor to a small value like 0.7 or even 0.5 in solution control.

Fanadam March 19, 2003 19:11

Re: A STRANGE PROBLEM!!!
 
Hi, Kaiser, Ming and Mateus, thank you very much for your advice. I've changed the under relaxation factor, the discretization number of momentum and even the residuals, but nothing can make contribution. really, I can not image!!!

kaiser March 20, 2003 02:54

Re: A STRANGE PROBLEM!!!
 
my final suggestion is to remodel ur problem, and provide more resonable grids.

Chetan Kadaki March 20, 2003 03:28

Re: A STRANGE PROBLEM!!!
 
First lets shed some light on what is happening. For each cell and each iteration, values are being calculated. After each iteration, the values are recalculated. When the values gradually come to a set of unchanged values, the solution is converged.

But sometimes, with all the equations and unknowns, and if the problem is complicated, the stability of the solution is in question and the solution may diverge. If the residuals are getting higher and higher, or if you find that a variable value at a point (e.g. pressure at 0,1) does not reach a consistent value, it is diverging.

This is why we have solution controls. When you decrease the under-relaxation factor, you are saying that the change in solution with each iteration will change by only a partial of what it is calculated to (for example if the calc says add 10 Pascal to that cell, but your u-r factor is .5, then the change will only be 5 Pascal). Simply, the u-r factors limit the chances that the solution will overcorrect itself. its like when you throw a ball against a wall. If you throw it too hard, you might not be able to catch it and it will bounce back and fly past you. In your case change your u-r factors to maybe .2 for pressure, .5 for momentum, k and epsilon, and .8 for other variables.

Also run it with 1st order discretization factors, and SIMPLEC. These are simpler equations that will remove the solution complication.

If the Courant number is in play, dcrease that too. The Courant number is a time stepping factor... the higher it is the faster the solution can be solved, but the less stable is the solution.

You also need to check the initial solution. How do you choose your initial conditions?

If it works for 600 m/s at inlet, run it for a few hundred iterations, and then change the inlet b.c. to 700 m/s and let it converge for another few hundred iterations, repeat for 800, 850 and finally 900 m/s. The solution can handle gradual change.

Try to visualize why your solution is diverging and then try to understand how to correct for it.

Another suggestion is to look at your mesh design. Adapt gradients, but a too fine a mesh can introduce more calcs and more potential problems.

Best of luck; hope I have helped

Kaiser April 13, 2003 14:58

Aerothermal example
 
Is there any good example of any aerothermal model.

I need to solve a transient energy equation first.

Then keep the result from the first run and then run the k-e turbulence model steady state and then do a transient thermal model again.


All times are GMT -4. The time now is 23:06.