|
[Sponsors] |
February 2, 2011, 13:44 |
Fluent Basics
|
#1 |
Member
John
Join Date: Jan 2011
Posts: 69
Rep Power: 15 |
Hi Guys, just to say this forums is too useful..
I have some question, let the contest start to you all. Try answering any question you can.. FEEL FREE. I am testing car aerodynamics in Fluent:- 1- How to perform GIS, I know that you have to keep editing your mesh. BUT my question is how to know that its OK. 2- How to do a 2nd order discretization. all I know is that you have to change the solution method (pressure and momentum) to 2nd order, but how to know that your answers are really in 2nd order and to compare it. 3- Drag and Lift (print to console and plot) forces are both ticked in monitors, but where can I see the actual answers of drag and lift. 4- I need to do a comparison between both 1st order and 2nd, how can I do that. And, if you have any car tutorial model from your college, university. Please upload it, it will be very helpful to me. Thanks a lot, John |
|
February 2, 2011, 14:18 |
|
#2 |
Member
John
Join Date: Jan 2011
Posts: 69
Rep Power: 15 |
For Q3:-
I know that my solution will be full converged if the the plotted coefficient remains constant while iterating, my plotted coefficient is not constant now what I am supposed to do in this situation? |
|
February 2, 2011, 18:59 |
|
#3 |
Member
John
Join Date: Jan 2011
Posts: 69
Rep Power: 15 |
no one in the whole thread? no experts or professionals here?
|
|
February 2, 2011, 20:43 |
|
#4 |
Senior Member
Real Name :)
Join Date: Jan 2010
Location: United States
Posts: 192
Rep Power: 16 |
By plotted coefficient, do you mean the residual? If so (perhaps even if not), be careful. Just because the residual remains constant does not necessarily mean the solution is fully converged to a physically realistic answer. Pick the physical parameter you're looking at and ensure that it remains constant to within the error tolerance you're OK with. For instance, in incompressible flow in a pipe w/o mass generation, perhaps you'd want to monitor the mass flow into and out of the system to ensure that they sum to zero. In fact, there are many times I un-check"check convergence" in the residual monitors, and simply watch the fluxes to ensure convergence. This is a very simple case, but can prevent significant rework in the future on a converged, but physically unrealistic answer.
If you're dealing with a parameter that simply won't "settle down" -- that is, it appears to sinusoidally fluctuate, do the following: 1) Review your boundary conditions to ensure they're valid and not inherently transient. The same goes for your physics. Multiphase slug flow has a natural "transient" quality to it which makes convergence to a rock-steady value difficult. 2) Review your mesh to ensure that it's of sufficient quality to capture the physics you're looking for (did you do a grid dependence check?) 3) Ensure your UDF's are behaving properly over the range of temperatures, pressures, etc. that the solver might be trying 4) Change the limits on your solve to maintain "realistic" solver attempts (especially true if you're using udf's -- see #3). For instance, if I have water at 300K flowing into a pipe which has wall temperature of 300, I don't need to have a 5000K upper limit on my temperature. The same goes for pressure. 5) Turn down your under relaxation factors a bit. Don't let the "rapid convergence" fool you, however. Dialing them down will give you a slower, but often more reliable approach to convergence with fewer bumps in the road...It also will often trip the default "convergence" levels (1e-3, etc.) in Fluent. ComputerGuy |
|
February 2, 2011, 20:47 |
|
#5 | |
Senior Member
Real Name :)
Join Date: Jan 2010
Location: United States
Posts: 192
Rep Power: 16 |
For 2 and 4, you're doing it correctly. To compare, pick a value you're interested in...For instance, pressure or drag on a face. From there, you can simply run two simulations, plot/print/write down or otherwise capture the data you're looking for. As far as I know, Fluent doesn't really have something that will automate the process of solution method comparison. When people to model validation studies (on grid refinement, for example), they are truly making 2,3,4,etc. different grids of various quality, running each simulation, and checking to see when the results stop changing.
ComputerGuy Quote:
|
||
February 3, 2011, 16:47 |
|
#6 | |
Member
John
Join Date: Jan 2011
Posts: 69
Rep Power: 15 |
Quote:
Is my mesh converged and can be used for 2nd order discretization? |
||
February 4, 2011, 02:11 |
|
#7 | |
Super Moderator
Maxime Perelli
Join Date: Mar 2009
Location: Switzerland
Posts: 3,297
Rep Power: 41 |
Quote:
Discretization order is independant from your grid (so far I know )
__________________
In memory of my friend Hervé: CFD engineer & freerider |
||
|
|
Similar Threads | ||||
Thread | Thread Starter | Forum | Replies | Last Post |
Stopping a Fluent batch job AND saving the data! Possible? | Volker Pawlik | FLUENT | 13 | December 28, 2020 04:16 |
Two questions on Fluent UDF | Steven | Fluent UDF and Scheme Programming | 7 | March 23, 2018 03:22 |
Compared MRFSimpleFoam and Fluent in a centrifugal pump! | renyun0511 | OpenFOAM Running, Solving & CFD | 8 | July 6, 2010 06:24 |
Fluent 12.0 is worst then Fluent 6.2 | herntan | FLUENT | 5 | December 14, 2009 02:57 |
Problems in lauching FLUENT | Lourival | FLUENT | 3 | January 16, 2008 16:48 |