| Author | Message | |||
     Chris Bailey (c_bailey) Member Username: c_bailey Post Number: 6 Registered: 01-2010 |
It seems like the new INACTIVE(some variables) feature is incompatible with the CONTACT = JUMP(expression with other variables) feature. If equations are only active inside a region whose boundary is a jump, the message BAD Matrix Reference (...) appears. Or, at least it seems to in the attached demonstration code, which gives slight natural convection in a square cavity (this was modeled after the sample buoyant.pde), the cavity being inside a larger region of conductive material without convection. This demo works if the CONTACT line is commented out but gives the error if it isn't. These two features should work together, right? I mean, they are physically reasonable, right?
| |||
| Marek Nelson (mgnelson) Moderator Username: mgnelson Post Number: 181 Registered: 07-2007 |
Yes, these two features should work together, as long as the CONTACT is being applied to a variable that is active on both sides of the boundary. Which it is in your script. We will have to look into this and get back to you. | |||
| Robert G. Nelson (rgnelson) Moderator Username: rgnelson Post Number: 1326 Registered: 06-2003 |
We have located some indexing errors when CONTACT and INACTIVE are combined, and we will correct these in the next release. However, it does not seem to me that either is necessary in this problem. You have a very low contact resistance in the temperature equation (which causes some numeric troubles), but even at 100 times this resistance there is no visible offset in the wall temperatures. This means that you could eliminate the wall altogether, which means you don't need to inactivate the velocities. | |||
| Chris Bailey (c_bailey) Member Username: c_bailey Post Number: 7 Registered: 01-2010 |
Thank you. This model does not need either of these features, I agree. It was only a demonstration in which they are physically reasonable but not quantitatively important. I am not permitted to post my real problem, where it does matter, and which takes a quarter hour to solve. In fact, I am always posting demonstrations rather than my real problems. I envy everybody who does not need to create demonstrations to post! | |||
| Robert G. Nelson (rgnelson) Moderator Username: rgnelson Post Number: 1327 Registered: 06-2003 |
PS You have the wrong sign on your CONTACT BC. This makes the solution unstable on the boundary. And it solves better with a higher resistance. My modification of your script is attached.
|
Demonstration of incompatibility