DaFoam Questions

[Janfoam]

Hi Ping He,

1. For the wind turbine tutorial, I noticed that no constraints were applied. Does that mean that only unconstrained simulations can be done in this particular case?

2. "The solveCL computes the angle of attack to match the target_CL by first computing dCL/dAlpha (alpha is the angle of attack). Then, it changes alpha using this formulation: Alpha_New = Alpha_0 + dCL/dAlpha * (CL_target - CL_0). We need to repeat this process until we get good CL_target." The changing of the alpha with the formulation Alpha_New = Alpha_0 + dCL/dAlpha * (CL_target - CL_0) - to be sure Alpha_0 is initial angle of attack it calculated while CL_0 is also the initial angle lift coefficient? Can you please clarify the reasoning behind using the formulation?

Thanks

[friedenhe]

1. Yes, we do not have any constraints for the wind turbine case at this moment. However, you are welcome to add thickness or volume constraints, refer to the Onera_M6 case on how to add these geometry constraints.

2. Yes, CL_0 is the initial lift, alpha_0 is the initial angle of attack. The solveCL function assumes the CL vs alpha relationship is linear. In that formulation, dCL/dAlpha is the gradient of CL wrt alpha, computed by using the finite-difference method. Once you know the gradient of the linear CL-alpha function, you can predict CL at any alpha. We need multiple iterations because the CL-alpha relationship is not very linear if CL_0 is far away from CL_target.

[Janfoam]

Thanks for your answers Friedenhe.

I looked through the code for the NREL6 and have some more questions:

• Why is symmetry planes specified as: symmetryPlanes": []?
• No CL value is specified at the start( usually global parameters). What is it using for the commands for Task to arrive at a CL?
  elif args.task == "solveCL":
  optFuncs.solveCL(CL_target, "alpha", "CL")
• There is no genAirFoilMesh.py in the case folder. Is this correct?

For the Onera6 case
Unfortunately, this aspect is still not clear to me. It is also in the airfoil optimization case. Can you please clarify what you use to define the lines in this case. They could be any lines that are close to the le and te, but how do you ensure that the lines are close to the leading and trailing edge? What is the guiding compass in this case?

NOTE: the LE and TE lists are not parallel lines anymore, these two lists define lines that

are close to the leading and trailing edges while being completely within the wing surface

leList = [[0.01, 0.0, 1e-3], [0.7, 0.0, 1.19]]
teList = [[0.79, 0.0, 1e-3], [1.135, 0.0, 1.19]]

[Janfoam]

The large words in bold are coming directly from the code. Not sure why they became so large and bold - this was unintentional :-)

[friedenhe]

@Janfoam There is no symmetry plane for the NREL6 case. The objective function is torque. We don't use the genAirFoilMesh.py to generate mesh, instead, we use snappyHexMesh. Read this page on how to setup leList and teList.

[Janfoam]

@friedenhe, Thanks for your answers. I'll be testing the simulations and will soon be in touch with other questions. Enjoy your Holidays.

[Janfoam]

Hi He,

-Can you please elaborate how the boundary conditions are defined in DAFoam. I've haven't come across any explanation for this in any of the publications associated with DAFOAM.

• How does DAFoam consider the velocity components when varying the angle of attack?

• For the FFD points, is there a criteria in DAFoam that should be respected?

• For thickness changes- is there an output file to look at to see how the thickness varies after the simulation? From the Constraint setup in DAFoam, the thickness constraint depends on the le and te, which are vary depending on how the user established the lines used to define the te and le.

Thanks

[friedenhe]

1. Check line 127 from here. We defined the x and y components of velocity at far-field based on the alpha value.
2. The only requirement is that the FFD points should be close to the design surface.
3. The constraint values will be printed to the screen after each flow simulation. Check this.

[Janfoam]

Thanks

For the boundary conditions I meant how is this determined mathematically? I am comparing DAFoam's discrete adjoint mathematical statements with the continuous adjoint. In the latter, there is a step by step guide as to how to arrive at the boundary conditions, however, I have not yet found that for DAFoam. Can you please clarify?

Also, is there any possibility to have online workshops/tutoring with DAFoam and if yes who should I contact for this?

[friedenhe]

In the discrete adjoint, we don't need special treatment or derive extra formulations for boundary conditions, as opposed to the continuous adjoint. The only thing we need to do is to update the boundary values before evaluating the residual. Check our paper for details. We do have plans to host workshops, maybe sometime later this year. We will post an announcement when the time comes. Stay tuned.

[Janfoam]

Hi,
I set up a simulation in DAFoam to simulate an airfoil with the aim of seeing how the thickness parameters work. I don't know if this is normal, but it took DAFoam around 4 hours to calculate the angle of attack that I will use for the simulation. The simulation is running, but its somewhere close to 17 hours already. I think it will take maybe another 10 or so to finish. Is this usual?

For the thickness constraints I noticed that both the trailing edge and leading edge are coded to be fixed in DAFoam and it is not recommended to change this because of the symmetry constraint. However, for a wind turbine while the trailing edge is usually modelled fixed the leading edge is usually free. My question is what can be done in DAFoam for this seeing there is a restriction in the thickness conditions?

How does the FFD points and movement affect the thickness in DAFoam?

[Janfoam]

I know that if the native OpenFOAM solver simpleFoam is not accessible then DAFoam will not run. There were no issues with the native solver nor DAFoam prior to running the simulation for NACA 0012. It is very unusual that after completing the simulation that these issues are occuring.

[friedenhe]

I suggest you first learn the basic usage of OpenFOAM from here, make sure you can run native OpenFOAM, and know how to deal with OpenFOAM-related errors. For example, this one: "cannot open case directory "/data/j/Correctmeshcopy/dafoam-2.1.3/tutorials-master/NACA0012_Airfoil/incompressible/processor0" is a very basic error in OpenFOAM. Google how to fix it.

[Janfoam]

Hi @friedenhe, Thanks for your quick response.
Some points:

• I do know the basics of openFOAM

• The "cannot open case directory" is not that simple to fix as outlined here and could require administrator privileges: https://www.cfd-online.com/Forums/openfoam-solving/134568-foam-fatal-io-error-cannot-find-file.html

• Even after the native solver loaded the error with DAFoam continues. It would be helpful to know why this happens so that I can move on with my project. If you are able to provide some insight into what is causing this error, this would be greatly appreciated.

• 

[friedenhe]

You managed to run solveCL before, and now it is not working. It is hard to tell what have you changed that caused this. So I suggest go back to the last working setup (it could be the tutorial case) and gradually change things, see which step breaks. If you can not run solveCL for the tutorial case. Check your DAFoam installation.

[Janfoam]

Thank you @friedenhe for your response and suggestions. I'm still waiting for the administrator to have a look at DAFoam installation after doing the necessary checks myself and could not find the problem.

In the meantime, to not lose momentum in learning about DAFoam, I'll ask some additional questions under different heading on the platform. Subject-specific headings will make it easier for people in the future to find what they want instead of scrolling through a long chat. Will come back to this thread once the issue is resolved.