|
[Sponsors] |
September 20, 2011, 09:09 |
CFD Sensitivity Check Method
|
#1 |
New Member
Join Date: Sep 2010
Posts: 28
Rep Power: 0 |
I know it's important to do a sensitivity check with grid refinement or turbulence model, but which do I do first? Should I:
1) Stick to a turbulence model first, then refine the mesh, then compare each turbulence model suitable at that mesh setting? 2) Or refine the mesh individually for each turbulence model chosen, then compare them? I've read the Wiki on Accurate results and "Why won't it converge", this is mostly more of a theoretical methodology-based question |
|
September 20, 2011, 09:16 |
|
#2 |
Senior Member
Andrew
Join Date: Mar 2009
Location: Washington, DC
Posts: 209
Rep Power: 18 |
I always do a grid check. Also, it is important to know that it is not always a 'refinement' that you are after. I usually do a 'double' rule in which I will cut the number of cells in half and see how my answer compares. I will also double the number and see how it compares. Sometimes you will find that you have way too many cells and you can speed the computational time by trimming that down. That might only be the case for GIANT grid numbers, so I guess it depends on how many you have to begin with. But, you do have to show that you have the mesh correct.
|
|
September 20, 2011, 09:30 |
|
#3 |
New Member
Join Date: Sep 2010
Posts: 28
Rep Power: 0 |
Yes, I always do a grid check by doubling the number of nodes, then checking the results, then doubling again if the results change a lot. But I'm wondering which I should do first, like I said above.
|
|
September 20, 2011, 09:33 |
|
#4 |
Senior Member
Andrew
Join Date: Mar 2009
Location: Washington, DC
Posts: 209
Rep Power: 18 |
I'd do grid first - if your mesh is bad then it won't matter what kind of turbulence model you pick.
|
|
September 20, 2011, 20:40 |
|
#5 |
New Member
Join Date: Sep 2010
Posts: 28
Rep Power: 0 |
All right. That's good. Thanks!
|
|
September 26, 2011, 07:04 |
|
#6 |
New Member
Join Date: Mar 2010
Posts: 25
Rep Power: 16 |
Hi, I've done quite some research on that for flow over different airfoils. My conclusion is that since different turbulence models can potentially result in different flow features, you should strictly speaking do a mesh study on every model that you intend to use. Now since that is a lot of work I found that similar simulation results can be obtained for each group: all one equation models, fully turbulent 2 equation models and 2 equation models with wall damping functions. Different transitional models I found behave quite different to one and another.
Also, I know that it may sound contradictory but don't always expect to get an asymptotic curve of a certain variable as you increase/decrease your mesh density. For instance a very fine near-wall mesh may result in divergence from the asymptotic region. Another example is the SA model which seemed to produce more scattered results than say for example the RNG k-epsilon model. Hope that helps. |
|
September 26, 2011, 10:11 |
Hello all
|
#7 |
Member
R. P.
Join Date: Jul 2010
Posts: 73
Rep Power: 15 |
Hi all,
I am new in the CFD area and I have been reading a lot of things that is necessary to spend a good time preparing a good CFD mesh. A good mesh will facilitate the code convergence and will give me accurate results. I also ready that the CFL number should be less than 0.2. In this way, which parameters should I take into account to generate a good mesh. Let's take the convergent divergent nozzle example with 500 mm of length and initial velocity equal 10 m/s. |
|
|
|
Similar Threads | ||||
Thread | Thread Starter | Forum | Replies | Last Post |
CFD Salary | CFD | Main CFD Forum | 17 | January 3, 2017 17:09 |
Code for most powerfull FDV Method | D.S.Nasan | Main CFD Forum | 6 | September 4, 2008 02:08 |
Taguchi Method and CFD | Xavier | Main CFD Forum | 3 | April 18, 2008 07:10 |
CFD JOBS and Expected Salary.... | Noel Harrison | Main CFD Forum | 11 | November 22, 2000 07:15 |
ASME CFD Symposium, Atlanta, 22-26 July 2001 | Chris R. Kleijn | Main CFD Forum | 16 | October 2, 2000 09:15 |