| Author | Message | ||
     Kevin Ellwood (ikevin) Member Username: ikevin Post Number: 4 Registered: 06-2003 |
I am sorry for the mistaken subject in my previous post. For some reason, my browser keeps changing it to my very first post in this new forum. Below is my question. I have a reaction diffusion problem where one edge has a time dependent dirichlet BC. The value is computed by assuming that the domain is in contact with a region that is well mixed. This means that the concentration in the well mixed region will change with time; hence, so will the direchlet BC. The value of this BC is computed based on the time integral of the flux of the diffusing species through the edge in contact with the well mixed region. My problem is that I get an error message: Implicit dependences of variables on TIME_INTEGRALs are illegal. Use a scalar variable. I undestand the issues why this has occured. In the past with my own codes, I have used lagging in which I compute the integral up to the previous timestep. I don't think I can do this in flexpde. Are there any tricks to get around this? | ||
| Robert G. Nelson (rgnelson) Member Username: rgnelson Post Number: 10 Registered: 06-2003 |
The problem with an explicit coupling is that in some problems it can go unstable. FlexPDE is not able to construct an implicit coupling through a Tintegral, so we block its use except in reporting information. The solution is the one suggested by the diagnostic: use a Scalar Variable. FlexPDE is able to build implicit couplings, and instability is avoided. If your integral is K=tintegral(f), then it is also true that dt(K)=f. So declare K as a Scalar Variable, and give it the equation dt(K)=f and an initial value of zero. [As to the functioning of the bulletin board, I find its behavior perverse. It is driven by Discus software, and the new version is much more confusing than the old one.] | ||
| Kevin Ellwood (ikevin) Member Username: ikevin Post Number: 5 Registered: 06-2003 |
Thanks, I just came back to the board to point out that I found an example that will allow (I think) me to do it. It is "heaterti.pde" and, as you pointed out, defines the differential form of the problem. Thanks a bunch for your note, Kevin |