CFD Online Discussion Forums

CFD Online Discussion Forums (http://www.cfd-online.com/Forums/)
-   STAR-CCM+ (http://www.cfd-online.com/Forums/star-ccm/)
-   -   Convergence problems post impact[URGENT help needed] (http://www.cfd-online.com/Forums/star-ccm/84631-convergence-problems-post-impact-urgent-help-needed.html)

alastormoody11 February 4, 2011 04:29

Convergence problems post impact[URGENT help needed]
 
Hi,

I am simulating sloshing in a rectangular tank which is initially moving with a variable acceleration and hits a wall and then starts moving in the opposite direction.

I have successfully simulated the inital part of the problem using the body force approach, in the motion after impact a zero acceleration a constant velocity in the opposite direction.

I am trying to validate my simulation with a paper which has both experimental and numerical results obtained from FLUENTwhich uses a time step of 0.01.

The convergence diffculties were bound to arise after impact due to the sudden change in the direction, however the problem is that even after I have reduced the under-relaxation factors as well as the time step to a ridiculously low degree(I ran the simulation with time step of 0.001s and tried using a time step as low as 1e-7 to no avail) the solution fails to attain a respectable residual level even for a single time step.

If any of you have any suggestions, please reply.

As a last resort I gave a maximum limit on the number of inner iterations eqaul to 750, after which I got results with the highest residuals around the 0.1 mark, though the resemblance to the experimental results is uncanny, I plan to extract quantitative data from the simulation which will be an exercise in fultility unless the simulation is properly converged.

I have tried different grids, however since the experimental results published in the paper basically outline the results in form of images of sloshing, I believe I cannot move in the direction of a very coarse mesh.

Any help in this regard will be highly appreciated.

abdul099 March 22, 2011 09:52

Not sure whether I've understood your problem: You worry about the residuals not falling through the floor in a time step?
Don't give too much about the residuals, and more important, don't give too much about the absolute level of residuals. As they are normalized, you shouln't run it until the residuals are below a certain value. They just need to drop for some (2 - 3) orders of magnitude.


All times are GMT -4. The time now is 11:42.