Residual Theta and Phi
Residual Theta and Phi
I am looking at the transport of heat and moisture in a wall structure with Wufi 2D (no rain). The convergence analyzer informs me that convergence failed for one time step because Phi = 1e-4 > 5e-5 after 1500 iterations. Furthermore iteration was interrupted for 12 time steps because there was no further improvement after 750 iterations for phi (and also for theta). In the worst case phi = 1e-3 >> 5e-5. How can I asses this information. What does this mean for the simulation results? How is theta and phi calculated?
-
Christian Bludau
- WUFI SupportTeam IBP
- Posts: 1161
- Joined: Tue Jul 04, 2006 10:08 pm -1100
- Location: IBP Holzkirchen, the home of WUFI
- Contact:
Re: Residual Theta and Phi
Hello,
see here for more information:
http://www.wufi-forum.com/viewtopic.php?f=15&t=439
and here
http://www.wufi-forum.com/viewtopic.php?f=15&t=768
Christian
see here for more information:
http://www.wufi-forum.com/viewtopic.php?f=15&t=439
and here
http://www.wufi-forum.com/viewtopic.php?f=15&t=768
Christian
Re: Residual Theta and Phi
Hello,
no further improvement: residual remains constant - it should decrease. Perhaps because of "oscillation"
residual phi: abs(max(R.H.-R.H._previous-iteration))
residual theta: abs(max(Temperature-Temperature_previous-iteration))
The convergence analyzer is just a graphical view of the iterative process
Usually the first iteration fails (because the initial conditions are unrealistic/constant across component) - you can safely ignore this
Interruption with message "No further improvement" results in a more or less inconsistent solution: Current temperature field would result in a different humidity field and vice versa without coming to an converged end. This can happen because of strong coupling of temperature and R.H. WUFI first solves for temperature (applying the current R.H. for R.H.-dependent properties) and then for R.H. using latest temperature for temperature-dependent properties. These steps are performed until either convergence criterion (conv-crit) or max. number of iterations is reached.
Your worst case 1e-3 means the last performed iteration changed R.H. e.g. from 0.5 to 0.501 in one grid volume. One could think, okay when WUFI would perform 100 additional iterations it would end up with RH of 0.6 in that volume which would be really bad. But usually the solution osciallates in such a case (0.5->0.501->0.5->0.501....) for any reason
no further improvement: residual remains constant - it should decrease. Perhaps because of "oscillation"
residual phi: abs(max(R.H.-R.H._previous-iteration))
residual theta: abs(max(Temperature-Temperature_previous-iteration))
The convergence analyzer is just a graphical view of the iterative process
Usually the first iteration fails (because the initial conditions are unrealistic/constant across component) - you can safely ignore this
Interruption with message "No further improvement" results in a more or less inconsistent solution: Current temperature field would result in a different humidity field and vice versa without coming to an converged end. This can happen because of strong coupling of temperature and R.H. WUFI first solves for temperature (applying the current R.H. for R.H.-dependent properties) and then for R.H. using latest temperature for temperature-dependent properties. These steps are performed until either convergence criterion (conv-crit) or max. number of iterations is reached.
Your worst case 1e-3 means the last performed iteration changed R.H. e.g. from 0.5 to 0.501 in one grid volume. One could think, okay when WUFI would perform 100 additional iterations it would end up with RH of 0.6 in that volume which would be really bad. But usually the solution osciallates in such a case (0.5->0.501->0.5->0.501....) for any reason