Author |
Message |
Daniil Andrienko (daniil)
New member Username: daniil
Post Number: 2 Registered: 08-2008
| Posted on Monday, August 25, 2008 - 08:18 am: | |
Dear Mr. Nelson! I'm trying to simulate field of temperature in 2D axial geometry using heat transfer equation with joule source of heat. And this program (I mean only heat transfer equation + joule source) works good. Then I want to simulate convective flux in the area, where I've solved heat transfer equation. I wrote a script , which solves only Navier - Stokes equation and it works, but when I'm making the united script (I mean heat transfer eq. + Navier - Stokes) then there is an error about infinity. Help me please. Thank you!Heat transfer + Navier Stokes Full.pde (3.3 k) |
Heat transfer + Navier Stokes Full.pde (3.3 k) |
|
Daniil Andrienko (daniil)
Junior Member Username: daniil
Post Number: 3 Registered: 08-2008
| Posted on Monday, August 25, 2008 - 08:20 am: | |
You need the second Full.pde, not the first one ! |
Robert G. Nelson (rgnelson)
Moderator Username: rgnelson
Post Number: 1161 Registered: 06-2003
| Posted on Monday, August 25, 2008 - 01:52 pm: | |
1) Your viscous flow equations are inconsistent. You impose a velocity on inlet and outlet, but also require that both be at the same pressure. You can't have a velocity without a pressure gradient. 2) You have a very low viscosity fluid and shock it with instantaneous velocity jump at the inlet and outlet. An instantaneous change requires an infinitesimal timestep to track the history of the change. Your fluid equations are steady-state, but the velocity terms in the temperature equation require time tracking of the velocity changes. You should start the system with self-consistent initial conditions (zero velocity, for example), and then ramp the boundary velocities over a physically reasonable time. This would also require a reasonable THRESHOLD on the velocity variables. 3) There appears to be no temperature effect on the fluid flow, so you could simply TRANSFER the flow velocities from the "viscous" problem and TRANSFERMESH them into the "full" problem, eliminating the U,V and P variables. If you need later to incorporate some flow dependence on temperature, you could retain the U,V,P equations and use the TRANSFER'ed data as initial values. 4) Requiring monitors at 1e-5 intervals prevents the timestep from rising above 1e-5. Use "FOR CYCLE=1" of "FOR CYCLE=10" instead. 5) You should always MONITOR every variable, at least during the development of a model. How else can you see what is happening? 6) Your THRESHOLD for TEMP is wrong. THRESHOLD is the minimum variation of the variable that must be held to ERRLIM control (see documentation). THRESHOLD=520 says you don't care about temperature oscillations less than 520 units. 7) Unless the final temperature variation is large compared to 520, it would probably be wise to compute the temperature difference from ambient, rather than the absolute temperature.
|
|