effect of changing underrelaxation parameter
 

hi, friends,

i am working on unsteady phase change phenomena during freezing of viscius liquid , my solution is getting converge only for initial time step , to overcome that i am changing underrelaxation parameter in between the itereation , does that make any difference in the solution, plz help me out

with regards kishore rathore

Re: effect of changing underrelaxation parameter
 

It depends on how much you reduce your underrelaxation factors.

If you reduce them too much, you risk to obtain a false convergence, or an unphysical solution.

Can you post the values of underrelaxation factors you're using?

Hi :)

ap

Re: effect of changing underrelaxation parameter
 

Underrelaxing just means that it will take more iterations to reach the final, converged solution. But beware, it may take along time to reach that final converged solution. I usually monitor a few important parameters till they stop changing with further iteration, and wait until the residuals have flattened before I consider the solution converged.

Re: effect of changing underrelaxation parameter
 

It's true that, usually, underrelaxing means it takes more iterations to reach the final result.

But, if you underrelax too much, the variable may change too little. So it seems that variable satisfies your convergence criteria, but actually it just doesn't change enough to make the solver iterate more.

It's what I called "false convergence" in my first answer. Obviously, in this case you have unphysical result (for example, time steps go on, but your solution is not the actual solution at that time step).

Hi :)

ap