Rotor blade simulation using SU2 and MBDyn


I’m doing a FSI simulation of a rotorblade and I’m facing a problem with the mapping.
The structure of the blade is discretised as a beam (1D reference line) with MBDyn, while the Fluid domain is discretised with an unstructured mesh on SU2. I’m attaching an image to show better the blade discretisation and the mapping problem.
What I observe is that after 1/5 of a revolution, the simulation stops because progressively the airfoils are getting thinner and thinner and the Mach on the blade surface is too high (should be 0.6 and it is 1.4).
I’m using rbf-thin-plate-splines as I’ve already tried nearest-neighbour but the same problem arises.
Have you got any idea why this is happening?

Thank you.

1 Like

The mapping from 1D to 3D could give problems. You could try to define a 3D helper mesh in MBDyn. Similar as in the master thesis of Qunsheng Huang.

@JulianSchl is working on a quite similar project, not sure if he saw your post.

Hi @alicezanella,

I had a similar problem coupling OpenFOAM and MBDyn.
The problem is, that you cannot map a line onto a surface or volume.
Therefore my solution was to interpolate the beam nodes onto the airfoil surface.
With RBF and nearest-neighbour mapping you cannot really steer the mapping direction, but only the mapping quality.
I’ve drawn a little figure to illustrate the mapping below:


Hi @JulianSchl

Yes, that is the exact problem I’m facing.
My question is: in MBDyn do you set the External Force Element to communicate with preCICE? What nodes do you set here? The beam nodes or the ones you’ve created on the interface?
And between the 1D node and the extrapolated mesh do you use a mapping?


Yes, I am using the external structural force element which exchanges the node positions of the red nodes and its corresponding forces of the beam nodes with the MBDyn-Adapter.

Inside the Python script of the MBDyn-Adapter, I’m extrapolating the nodes onto the fluid mesh to obtain the “new” orange mesh. Since the beam element should represent the camber line, I’m just “moving” the nodes perpendicular to the beam onto the fluid mesh to obtain my extrapolated mesh (as you can see in the figure).
I’m not completely sure if this is a legal way and how the mapping would be called (maybe somebody could assist in verifying that here :slight_smile: )

Thank you for your explanation! :smiley:
So the Fluid solver is using the displacements from the orange nodes and applying the forces here. Then you interpolate and transfer the forces to the red nodes and back again.

I’ll try to do something similar in my adapter! Thank you!

Are you using a symmetrical airfoil?
Are you simulating an elastic blade or Rigid blade? Why do you have to map in this case.

In case you are using Rigid blade, no mapping is required.
In case you are simulating a elastic blade. Make sure to extract the elastic motion before passing it to CFD solver. You can use this output to see, how realistic is your beam deformation.

No, it is a NACA23012.

MBDyn is passing Displacement Deltas to SU2, that is the variation of position between two consecutive time steps.
I didn’t have problems doing the vertical flap test case and a 2D benchmark (Square with attached flexible cantilever). I suspect the problem is passing from 1D to 3D…

As @uekerman mentioned, please look to the thesis of master thesis of Qunsheng Huang.
For the thesis, we used also a tabbed NACA23012 airfoil.

We solved the problem by introduction a 2D layer to the coupling.
You can apply map the 1D deformation to 2D plane. Then map the 2D plan to the 3D surface.

Let me know if you have questions regarding the master thesis.