|
[Sponsors] |
August 4, 2009, 19:35 |
what to monitor besides residuals?
|
#1 |
New Member
Franz Roman
Join Date: Jun 2009
Posts: 28
Rep Power: 17 |
Hi, I am running simulations of simple geometries with one velocity inlet and one pressure outlet and a porous zone. I make some variations to a base geometry. my goal is to find a geometry that gives the best (most uniform) velocity distribution at the outlet. For some geometries the residuals dont drop lower than say 3e-4 for epsilon (I am using k-epsilon realizable) and maybe not lower than 1e-5 for the other residuals. They go flat at this point. For other cases the residuals go flat at a lower value of the residuals. Well, the residuals look flat in the graph for hundreds or even thousands of iterations, but the values change a little bit as seen in the solver main window, so little that in the graph there is a horizontal line. Is that flat enough? or flat means absolutely no change in the values of the residuals?
I have read that it is good to monitor some value of interest to judge convergence. However, in my case I dont know which value this could be. I am not so much interested in any forces or pressures. I can only think of the mass flux, or the mass flow rate at the outlet. Would this be a good monitor for my case? thanks in advance Franz |
|
August 5, 2009, 00:39 |
|
#2 |
Super Moderator
Maxime Perelli
Join Date: Mar 2009
Location: Switzerland
Posts: 3,297
Rep Power: 41 |
monitor the massflow rate, and iterate until you become also flat curve
__________________
In memory of my friend Hervé: CFD engineer & freerider |
|
August 5, 2009, 08:02 |
|
#3 |
New Member
Franz Roman
Join Date: Jun 2009
Posts: 28
Rep Power: 17 |
Hi mAx,
thanks for answering. A question though. Flat curves means that the residuals change so little around a value so that the curves "look" flat? or means that the residuals dont change one bit at all? What about the massflow rate? does it have to be absolutely flat (no change in the values at all with ongoing iterations)? or is flat just how it looks in the graph, but with possible minor changes from one iteration to the next? If the monitored quantities go flat, is it then ok if they do it at 1e-4 or 1e-5? or should they go flat at a lower level? many thanks Franz |
|
August 5, 2009, 08:23 |
|
#4 |
Super Moderator
Maxime Perelli
Join Date: Mar 2009
Location: Switzerland
Posts: 3,297
Rep Power: 41 |
*flat residuals curves mean that residuals don t converge anymore
*massflow rate curves should be flat, but not absolutely. Value can smoothly change, but if you still get oscillations, then it is not ok. It depends also on the precision you will reach. *For me a flat monitoring with all residuals below 1e-3, it's ok. (in my case)
__________________
In memory of my friend Hervé: CFD engineer & freerider |
|
August 5, 2009, 08:26 |
|
#5 |
Member
Join Date: Mar 2009
Posts: 32
Rep Power: 17 |
Hi
You can monitor mass flow rate at outlet, as well as total pr. at inlet as both the monitors show stabilization i.e. least variation and with your residuals fall in the plot(1e-3/4/5/6 your experience will tell you). You can conclude the run has formidably converged. |
|
March 21, 2013, 03:59 |
Monitor Plots
|
#6 |
Member
Thiagu
Join Date: Oct 2012
Location: India
Posts: 60
Rep Power: 13 |
Monitor plots are the important are primary parameters, very closely related with your objective of the problem.(which one wants to verify using CFD tool) .
In simple single phase flow problems 1. Mass in & mass out 2. area ave pressure # For Heat transfer problem 1. Min,Max ,Avg temperature Example: Air is heated by wall temp (imposed fixed wall temp BC), one can monitor (min/max/avg)temp of air at different location or in the region of interest, which will show solution fluctuation there. For steady state problem, it will be smooth. It is not necessary that it should be flat always, again it depends on physics at that location/local mesh refinement and calculation method (node avg/face avg) To me , monitor plots = Experimental data (real time test rig readings from probes) |
|
Thread Tools | Search this Thread |
Display Modes | |
|
|
Similar Threads | ||||
Thread | Thread Starter | Forum | Replies | Last Post |
[Tutorials] Tutorial of how to plot residuals ! | wolle1982 | OpenFOAM Community Contributions | 171 | February 20, 2024 02:55 |
Mesh size and solver residuals... | Scott | CFX | 5 | December 15, 2008 17:10 |
Convergence - scaled vs unscaled residuals | HS | FLUENT | 1 | November 7, 2005 05:45 |
residuals | Karl | CFX | 2 | June 24, 2003 23:13 |
Residuals? | Tom | Main CFD Forum | 6 | November 28, 2002 04:09 |