Please wait...
×

Error

Posted Fri, 15 Jun 2012 08:45:03 GMT by lemonok
Hi, everyone
I am confused in time control method in simulation  process.
For the borehole system, I tried two time control method, one is constant time step method. The simulation could continue until the end of the time, but the results is not good.
I changed to automatic time step control via predictor-corrector scheme , but it could not continue simulate after one day. (I have 600?days to simulate), the time increment could not continue, error is always bigger than the tolerance. (shown in attached documents,), I do not know what happened .
I want to know what is the difference for the results calculated by these two time control method?
one more, I do not know what is the problem when I simulate by the automatic time step, Could some one help me ?
Thank you a lot

Posted Mon, 02 Jul 2012 13:05:22 GMT by cassius
which timestep scheme you adopted?
I experienced some troubles with the Adams-Bashfort scheme, you could try with Forward Euler/Backward Euler...
Posted Tue, 03 Jul 2012 09:51:37 GMT by lemonok
Thank you for your help.
I tried with the constant time step.  and also I tried with variable time step  with Forward Euler/Backward Euler...
I found if I use the constant time step, in some time point, there are no convergence, but it continues to run to the next time point.  however, for the variable time step, if there is no convergence, the operation will not  continue until the convergence is done. if there is no convergence, the wrong operation happens.
In fact, I compared the results get from the constant operation model and variable operation model. there are seems not big difference.
So what I am confusing is how to choose the time step. in what case the constant time step should be chosen, and when the variable time step is prefered?
Thank you for your help.
lemon


Posted Wed, 04 Jul 2012 12:01:00 GMT by Denim Umeshkumar Anajwala
We do not use the contant time stepping option very often. As long as the time step is not chosen extremely short, it tends to 'jump' over important periods in the simulation, then possibly leading to non-convergence in some time steps. I would recommend to use an automatic scheme and check the model for reasons of instability instead.
Posted Wed, 04 Jul 2012 14:11:25 GMT by lemonok
Thank you very much peter,
You are right, I have found this problem that if I use the constant time step, some point  non-convergence occurs, so the calculation jumps which leads some wrong results.
I will recheck the model itself, and use the automatic scheme, thank you so much.

Lemon

You must be signed in to post in this forum.