Author |
Message |
Wang Yuan_peng (konggu168)
New member Username: konggu168
Post Number: 1 Registered: 02-2006
| Posted on Monday, February 27, 2006 - 01:12 am: | |
Hi Flex-users, I am solving a time-dependent diffusion problem in 1D as follows: Da*div(grad(Ca))-Da*alpha^2*(Ca-Cs)=dt(Ca) Where Da=1e-8,Cs=0.005,alpha=1e10,with intial and bondary conditions as follows: For t=0: Ca(x=0)=0.5,Ca(x>0)=0 For t>0: Ca(x=0)=0.5,dx(Ca) in the point x=1 is zero. I have tried a lot,but it gives the error message that its timestep has fallen below 1e-009 of start value.How can I deal with the intial and bondary conditions? Is the CONTACT b.c. applicable in 1D problems? Thank you for help! Wang |
Robert G. Nelson (rgnelson)
Moderator Username: rgnelson
Post Number: 553 Registered: 06-2003
| Posted on Monday, February 27, 2006 - 02:14 pm: | |
Simple arithmetic shows that your system has a relaxation time of 1e-12. You should specify an initial timestep of this order. FlexPDE defaults the initial timestep to 1e-4 times the specified time range. So if your specified time range is more than 10, FlexPDE will be forced to cut and re-cut the timestep until it is reduced by more than 1e-9, at which point the cutting process is halted. You have a discontinuous initial condition, which in principle requires an infinitely small timestep for accurate resolution. The initial smoothing pass used in FlexPDE will smear this over the size of the first cell, but it still may require a very small step size. |
Wang Yuan_peng (konggu168)
New member Username: konggu168
Post Number: 2 Registered: 02-2006
| Posted on Monday, February 27, 2006 - 10:32 pm: | |
Thank you very much!I am a new user of Flexpde,I will make an effort to solve problems. |
|