Oscillating Airfoil - Dynamic meshing or user-defined velocity profile.
Hello there esteemed CFD friends,
I am currently simulating a sinusoidal pitching NACA0012 airfoil in Fluent CFD using unsteady RANS. The two methods that I am considering is :-
1. Move the airfoil boundary as a rigid body and use remeshing and smoothing to deform/move the mesh according to the movement of the airfoil.
2. Keep the airfoil and mesh stationary and define a periodic boundary condition that will change the angle of attack of the inlet flow with respect to the stationary body at each timestep.
Now I have done method 1 and I seem to be able to get quite good results with it. Of course method 2 will be alot simpler and would apply better for large airfoil oscillations, but I am somehow skeptical about method 2 as I am not too familiar with the solver numerics.
Specifically my question is will the change in boundary condition flow at the start of each time step be convected throughout the computational mesh at the end of each time step. What I am worried here is that this is not the case and the flow solutions I am getting will somehow be lagging behind the movement of the airfoil.
I apologize if this is question is trivial, but I am trying to understand the numerics behind the two different methods better and I hope you CFD veterans will be able to shed some light for me on this matter.
Can someone at least point me to the correct literature if its too long and time consuming to explain it here?
Hi there again,
I have done some digging around and have found some studies regarding this. In general i have seen a number of ways people normally compute dynamic stall of airfoil problems.
1. Moving the whole mesh as a rigid body
(Flow Prediction Around an Oscillating NACA0012 Airfoil at Re 1000000 by Frederich, Bunge, Mockett and Thiele)
2. Using the sliding mesh method, where there is a nonconformal interface between two different fluid domains and one will rotate against the other.
(Hybrid Grid Approach to Study Dynamic Stall by Reu and Ying)
3. Moving the airfoil boundary as a rigid body and deforming/remeshing the mesh around it accordingly.
(Unsteady Euler Airfoil Solutions Using Unstructured Dynamic Meshes by Batina)
However there hasn't been a detailed discussion about what happens to the transport variables at each of the grid points when they move. Will they be interpolated?
Also there hasn't been a comparison of the differences of each of those techniques (if there is any). Is anybody familiar with the differences care to explain?
Method 1 is best if you are modelling a free field.
Method 2 is best if you have objects away from the airfoil, eg wind tunnel walls, ground, aircraft body.... etc.
Method 3 is computationally slowest and hardest to get working properly, but it can handle more general motions than the previous 2.
But really you should be able to get any of the methods to work - and you have found literature in all approaches so that shows it.
The differences between the method is best appreciated by trying them all yourself. Run them all and don't forget to do a sensitivity study on all important parameters to make sure you are sufficiently accurate. They should all give the same answer within a numerical accuracy.
I think you will find the reason nobody has published a comparison is because nobody has done one before - and here is your chance to do something nobody has done before. That is always pretty cool.
Thanks for the reply, greatly appreciated! By the way do you happen to know a good reference book that covers the numerics behind moving grids extensively? Im trying to understand the numerics behind it better and so far ive only come across a book by Ferziger and Peric with a small section in moving grids.
The numerics is best explained in the software documentation. Other then that you would have to look in the academic literature.
|All times are GMT -4. The time now is 19:18.|