|
[Sponsors] |
How to fix Mesh for the simulation to get accurate Lift and Drag |
|
LinkBack | Thread Tools | Search this Thread | Display Modes |
May 17, 2023, 13:21 |
How to fix Mesh for the simulation to get accurate Lift and Drag
|
#1 |
New Member
Namith
Join Date: May 2023
Posts: 6
Rep Power: 2 |
Dear Forum,
I am trying to simulate a rugby ball to generate the lift and drag forces and I am getting negative values for the forces and would request the forum on the way forward to fix the error. Computation power is limited and have to get results with coarse mesh. Base Size - 0.01 to 0.008 Target Surface Size - 0.01 to 0.008 Minimum Surface Size - 0.01 to 0.008 Surface Growth Rate - 1.2 Number of Prism Layers - 5 Prism Layer Stretching - 1.2 Prism Layer Total Thickness - 0.0343 Volume Growth Rate- 1.3 velocity - 20m/s |
|
May 17, 2023, 13:38 |
|
#2 |
Senior Member
Matt
Join Date: Aug 2014
Posts: 947
Rep Power: 17 |
Are you sure the report is set up correctly?
|
|
May 17, 2023, 14:00 |
|
#3 |
New Member
Namith
Join Date: May 2023
Posts: 6
Rep Power: 2 |
Hi,
Drag and lift Force Coefficient direction [1,0,0] & [0,1,0] density 1.284 velocity 20m/s area 0.00944293833 m^2 for ball part- region - ball |
|
May 17, 2023, 14:05 |
|
#4 |
Senior Member
Matt
Join Date: Aug 2014
Posts: 947
Rep Power: 17 |
that doesn't really answer the question... assuming your freestream velocity is along x, then your lift is acting in y and not z? are you calculating pressure and viscous forces or just pressure forces? is this 2D/3D, steady/unsteady?
If you want people to help, you will need to provide more context. |
|
May 17, 2023, 14:11 |
|
#5 |
New Member
Namith
Join Date: May 2023
Posts: 6
Rep Power: 2 |
Yes lift is acting in Y.
I am calculating the pressure forces only. it is 3d model and is steady state and i have used Spalart-Allmaras in the physics model. |
|
May 17, 2023, 14:15 |
|
#6 |
Senior Member
Matt
Join Date: Aug 2014
Posts: 947
Rep Power: 17 |
if you are only looking at pressure forces, then you might end up with negative drag depending on the pressure distribution and shape of the object. As for the negative lift, the shape and orientation can matter there too... for example, is this at a negative angle of attack? if you could post an image or two, especially one showing your flow field, that would go a long way.
|
|
May 17, 2023, 15:21 |
|
#8 |
Senior Member
Matt
Join Date: Aug 2014
Posts: 947
Rep Power: 17 |
Now we are getting somewhere... based on what you show, you have an angle of attack of +10 degrees not -10 degrees and in an ideal world would be getting a positive lift. However, your thickness/chord ratio is really high so it's possible that you have enough separation/stall that you trash your lift.
Have you checked your wall y+ values, are they ok? What is your wall model? Why spalart-allmaras, out of curiosity? |
|
May 17, 2023, 15:51 |
|
#9 |
New Member
Namith
Join Date: May 2023
Posts: 6
Rep Power: 2 |
Y+ is in the buffer region which is wrong.
I choose spalart-allmaras since it required less computation time compared to K-Epsilon which would require denser mesh. If getting the desired results with K-Epsilon is possible with bit coarse mesh would like to get suggestions on how to proceed with it since it is much more suited for Lift and drag simulation.Prism layers.jpg Y+.jpg |
|
May 17, 2023, 16:05 |
|
#10 |
Senior Member
Matt
Join Date: Aug 2014
Posts: 947
Rep Power: 17 |
I think your prism mesh has a lot to do with it. Try reducing the thickness so that the attached boundary layer stays contained, then adjust the thickness and stretching ratio so that you get a smooth transition into the core mesh. I would switch to the k-e model and use the default settings, then target a y+ around 30-60. Also, you could probably make your min surface size a little smaller to improve the resolution of the body surfaces with high curvature. If you have a lot of separation going on, you are probably better off moving to k-w SST and using a finer wall mesh, around y+=1. It handles adverse pressure gradients a little better than k-e.
|
|
May 17, 2023, 16:16 |
|
#11 |
New Member
Namith
Join Date: May 2023
Posts: 6
Rep Power: 2 |
I calculated the prism layer thickness using 0.37/(Re^(1/5)) and the stretching as 1.2. So should i reduce the thickness by a factor of 10?
What would be the minimum base size to K-e to get the desired output. |
|
May 17, 2023, 16:39 |
|
#12 |
Senior Member
Matt
Join Date: Aug 2014
Posts: 947
Rep Power: 17 |
A stretch of 1.5 is perfectly acceptable and probably gets you where you want to be a little quicker. I am guessing 0.37/Re^1/5 is some sort of flat plate BL function? To be honest, I rarely size it ahead of time. I will get something I think is appropriate based on inspection/intuition, run the solver then tweak the mesh based on the flow field that results to meet the guidelines I offered earlier. Not sure if a factor of 10x is needed, but you could certainly reduce the thickness some.
As far as a 'minimum base size to get the desired output'... I would simply say that is a difficultly posed question. I don't know what your 'desired' output is nor is there a minimum base size that gives you that answer, whatever it is. I suggest you conduct a mesh dependency study if you need your results to be independent of mesh settings. If you are unfamiliar, with mesh dependency, there are many threads on this topic already, so I would start there. |
|
May 18, 2023, 13:25 |
|
#13 |
Member
Join Date: Nov 2019
Posts: 93
Rep Power: 6 |
Accurate drag prediction for simple bluff bodies is actually pretty difficult, especially if you're limited to RANS. One example where you can see this is the flow around a smooth sphere. Different turbulence models will give you very different results and you might have a hard time matching experimental data (which are publicly available for the sphere so it's a good case for testing a model like this). I'm not saying your model is wrong, but just be careful.
|
|
May 22, 2023, 19:25 |
|
#14 |
Member
Kailee
Join Date: Dec 2019
Posts: 35
Rep Power: 6 |
One small thought... if you're compute bound (do you mean memory?) then the mesh cetainly deserves some optimization. Inlet and outlet shouldn't need the refinement you have on them. It's not a big win but if you're limited every little bit might help.
The domain looks on the small side if anything, certainly laterally I'm not sure you might influencing the results by the proximity of the walls (are they walls, or part of the inlet, or symmetry?). I would also encourage use of k-w SST here, it reduces the potential error within the boundary layer as stated already. I think you'll definitely benefit from a mesh independence study, or at least a study into the thickness of the prism layer. They're really not that much work if you set up your meshing properly with most refinements referencing correctly to a base size in a problem without much challenging geometry as in this case. But for sure RANS will be an issue here, as may be the (missing?) rotation of the ball that is such in important part of this problem. One more thing - I'm surprised by the "convergence" after as few iterations as in your posted images. It would be interesting to see what happens to the residuals after a few 100, not just a few dozen iterations... K. |
|
June 30, 2023, 19:36 |
Rugby Ball
|
#15 |
New Member
Mudassar
Join Date: Jun 2023
Posts: 1
Rep Power: 0 |
how can I find the angle of my ball 🏈. As it is tilted
|
|
|
|
Similar Threads | ||||
Thread | Thread Starter | Forum | Replies | Last Post |
Calculation of lift and drag coefficients on airfoil | CoolHersheys | OpenFOAM Post-Processing | 5 | September 27, 2021 06:04 |
How to not overwrite drag and lift coefficients after a simulation | Giovanni Trovato | FLUENT | 1 | August 1, 2018 00:31 |
Drag and lift calculation in Arbitrary Mesh Interface simulation | anon_q | OpenFOAM | 7 | April 22, 2018 13:24 |
wrong SU2 calculation for lift and drag coefficient for NAC4421 | mechy | SU2 | 7 | January 9, 2017 05:18 |
Mesh Grid Study - Result Tolerance - Lift, Drag, Moment. | Wingman | Main CFD Forum | 4 | November 14, 2016 16:50 |