|
[Sponsors] |
Convergence criteria: arithmetical average of imbalance |
|
LinkBack | Thread Tools | Search this Thread | Display Modes |
December 20, 2022, 10:03 |
Convergence criteria: arithmetical average of imbalance
|
#1 |
New Member
Join Date: Sep 2022
Posts: 3
Rep Power: 3 |
Hello,
for my simulations I have tried to reach satisfactory convergence by setting the conservation target to 0.002 (to reach Imbalance < 0.2%). After the run I checked the imbalance plot in the solver manager and saw the following (see attached picture). As you can see, the imbalances has indeed reached 0.2% which corresponds to the target. Nevertheless, the imbalance is "bouncing" and slowly converging which means we are more around 0.6% (as it was at around 300 steps). Now my approach would be to set the arithmetical average of the imbalance to be under 0.2%. Another option could be the standard deviation but this configuration would probably take more time to reach the conservation target. My question(s): is this a reasonable approach? How can I implement it in CFX-Pre? I know about expressions but I'm missing the knowledge of which function/expression expresses what I need. I am using CFX Pre (Ansys v2022.R1) for a compressor stage (rotor + vaned diffuser) simulation. The set up has already converged with satisfactory results but requested monitoring from my side. Now I want to make the simulation more autonomous. I have managed to automate the timestep and now I'd like to set the convergence properly. Thank you for your help! |
|
January 2, 2023, 01:04 |
|
#2 |
Super Moderator
Glenn Horrocks
Join Date: Mar 2009
Location: Sydney, Australia
Posts: 17,697
Rep Power: 143 |
It would be better to just get it to converge. Taking averages of imbalances versus iteration number has no physical justification I can see. See FAQ: https://www.cfd-online.com/Wiki/Ansy...gence_criteria
__________________
Note: I do not answer CFD questions by PM. CFD questions should be posted on the forum. |
|
January 2, 2023, 13:59 |
|
#3 |
Senior Member
Join Date: Jun 2009
Posts: 1,800
Rep Power: 32 |
Contact Ansys CFX support, request advice to measure convergence, and automatically stop the simulation.
There is the approach of using the Coefficient of Variation <N> statistic on a significant quantity of interest. CoV = Standard Deviation / | Mean Value | where the numerator and denominator are computed over the last N iterations. Usually, N is @50, which is CoV-50. Using the latest release of Ansys CFX, you can create a monitor expression for the quantity of interest, activate the monitor statistics Coefficient of Variation, and then set the number of iterations to be used. You can similarly compute such quantity after the fact, i.e. during the simulation using the Ansys CFX Solver Manager by adding a Derived Variable and selecting Coefficient of Variation as well. Unfortunately, you must manually stop the run this way
__________________
Note: I do not answer CFD questions by PM. CFD questions should be posted on the forum. |
|
Thread Tools | Search this Thread |
Display Modes | |
|
|
Similar Threads | ||||
Thread | Thread Starter | Forum | Replies | Last Post |
Convergence | Centurion2011 | FLUENT | 48 | June 14, 2022 23:29 |
ANSYS Mechanical - change convergence criteria | KeganLeckness | Structural Mechanics | 2 | May 12, 2022 05:57 |
Residual transient case, convergence criteria | Ursmooth | Main CFD Forum | 1 | September 30, 2017 03:18 |
Out File does not show Imbalance in % | Mmaragann | CFX | 5 | January 20, 2017 10:20 |
CAVITATION (convergence criteria & discretization) | ROOZBEH | FLUENT | 1 | October 6, 2003 09:32 |