Author |
Message |
Zhe Zhang (zhezhang)
New member Username: zhezhang
Post Number: 1 Registered: 05-2005
| Posted on Wednesday, May 11, 2005 - 03:01 pm: | |
Hello, Robert, I am a new user of flexpde and I have a problem in solving convective mass transport equation. Without the mass transport equation, u, v, p equations can be convergenced easily. However, it cannot get a convergenced solution with the mass transport equation. Attcahed is a simple test script for solving a fully-developed mass transport in a circular pipe. Can you tell me what the problem is. Thanks.
|
Robert G. Nelson (rgnelson)
Moderator Username: rgnelson
Post Number: 359 Registered: 06-2003
| Posted on Wednesday, May 11, 2005 - 06:59 pm: | |
1. FlexPDE scales all error measures for each variable to the greater of: a) the observed range (max-min) of values of the variable, or b) the declared threshold. In your C equation, you have a value condition of 1.0 at one end, and reflective boundaries everywhere else, so the solution is a uniform C=1 everywhere, and the observed range goes to zero. You have no threshold statement, so all errors are measured relative to zero. The convergence condition is then delta(C)/0<ERRLIM. This is a prescription for infinite computation. Since there is no variation of C in the problem, you must provide a threshold statement telling how much variation you care about. What is the scaling for the errors in C? 1? Then say so. It appears that V has the same problem. It is zero everywhere. (See THRESHOLD in the Help Index). 2. Your request for ERRLIM=1e-15 is wildly optimistic. One part in 1e15 is close to the machine precision for a single operation. You'll never get it over an entire mesh approximation. Try the default ERRLIM=0.001. If you're really anal, you can cut it by a few decimal places.
|
|