# laminar Flow over a sphere(laminar vs KOmegaSST simulation)

 Register Blogs Members List Search Today's Posts Mark Forums Read

March 12, 2023, 01:50
laminar Flow over a sphere(laminar vs KOmegaSST simulation)
#1
Senior Member

Join Date: Nov 2019
Posts: 205
Rep Power: 7
Hello all
I am simulating flow over a sphere at Re = 881 where Drag coefficient must be almost Cd=0.5. I simulate it using both using laminar and Turbulent models(KOmegaSST). for laminar simulation, I get Cd ~0.5 but fort the Turbulent simulation, first of all it takes very long tome to converge and even after convergence, it gives Cd~0.65(Actually it has not fully converged yet, see attached figure). Now I have two questions;
1- what happens if I use turbulent model to simulate laminar flow?
2- when should I expect convergence based on the attached figure?

Thanks,
Attached Images
 photo_2023-03-12_00-39-10.jpg (103.0 KB, 37 views)

March 15, 2023, 22:01
Sphere in Re-881 with laminar and kOemgaSST(SimpleFoam)
#2
Senior Member

Join Date: Nov 2019
Posts: 205
Rep Power: 7
I did test with laminar and kOmegaSST for sphere and airfoil and as I told earlier it predict higher drag Coefficient, but flow field visually seems correct. I repeat it with airfoil with angle of attack = 30 degree and not only drag coefficient is different, but also flow field is visually seems incorrect, why?

Also, I want to test below methods too, but they fail at the very beginning of the simulation;
1) SpalartAllmaras (fails at the beginning iterations),
2) kOmegaSSTLM (fails at the beginning iterations) ,
3) kOmegaSSTSAS (Selecting LES delta type vanDriest Killed)

Should I change my boundary conditions when I switch from kOmegaSST to other models?

Thanks,
Attached Images
 Spheree.jpg (45.3 KB, 31 views) Airfoil.jpg (80.1 KB, 21 views)

March 20, 2023, 06:49
#3
Senior Member

Join Date: Dec 2019
Location: Cologne, Germany
Posts: 363
Rep Power: 8
Quote:
 Originally Posted by farzadmech Hello all I am simulating flow over a sphere at Re = 881 where Drag coefficient must be almost Cd=0.5. I simulate it using both using laminar and Turbulent models(KOmegaSST). for laminar simulation, I get Cd ~0.5 but fort the Turbulent simulation, first of all it takes very long tome to converge and even after convergence, it gives Cd~0.65(Actually it has not fully converged yet, see attached figure). Now I have two questions; 1- what happens if I use turbulent model to simulate laminar flow? 2- when should I expect convergence based on the attached figure? Thanks, Farzad
1. one should not use a turbulence model when the flow is laminar. turbulence models use special boundary conditions to mimic real turbulence behaviour at walls. that will lead to wrong pressure and turbulent viscosity and thus to wrong drag coefficients for laminar cases.
2. for convergence you should always monitor the value of interest, in your case the drag coefficient. if it does not change or only oscillates between two stable extremes, you can say that based on your case setup that value is your solution.

March 20, 2023, 07:00
#4
Senior Member

Join Date: Dec 2019
Location: Cologne, Germany
Posts: 363
Rep Power: 8
Quote:
 Originally Posted by farzadmech I did test with laminar and kOmegaSST for sphere and airfoil and as I told earlier it predict higher drag Coefficient, but flow field visually seems correct. I repeat it with airfoil with angle of attack = 30 degree and not only drag coefficient is different, but also flow field is visually seems incorrect, why? Also, I want to test below methods too, but they fail at the very beginning of the simulation; 1) SpalartAllmaras (fails at the beginning iterations), 2) kOmegaSSTLM (fails at the beginning iterations) , 3) kOmegaSSTSAS (Selecting LES delta type vanDriest Killed) Should I change my boundary conditions when I switch from kOmegaSST to other models? Thanks, Farzad
you should always check if your boundary conditions are ok with the chosen turbulence model and also with your mesh sizing at the wall. for example, you should not expect good solution when your y+-value is 1 and you use k-Epsilon, which is valid only for y+=30 at least.
i think to predict good pressure distribution for forces around bodies you want to resolve the wall boundary layer, check your BC for your chosen turbuluence model if it supports boundary resolution.

if you have problems with convergence, first start your simulation with first order upwind schemes for advection terms: div(phi,u), div(phi,k) etc.
you can also use higher viscosity values.
after that try using lower viscosity with the previous converged solution. after that switch your div(phi,u) to second order upwind, and not change turbulence schemes.
once converged change your schemes for turbulence to second order scheme also.

try this step by step approach, do not rush to find a perfect solution right from your first try.

 Tags komegasst, laminar, openfaom, sphere