Cpmin constraint in airfoil optimization

[Roehrich]

Hi,

I was playing around with airfoil optimization and the speed an results of this code are amazing.

I would like to add a cpmin constraint. I was using the tutorial example

tutorial/06 - Aerodynamics/02 - AeroSandbox 2D Aerodynamics Tools/02 - NeuralFoil Optimization.ipynb
as a base.

I simply added

CP_min_constraints = np.array([-8, -8, -8, -8, -8, -8])

at the top (high values for testing).
and added my contraint in

opti.subject_to([ aero["analysis_confidence"] > 0.90, aero["CL"] == CL_multipoint_targets, np.all(aero["Cpmin"] >= CP_min_constraints),...

I get the following error

File "C:\Portable\Python\Neurafoil\lib\site-packages\aerosandbox\optimization\opti.py", line 482, in subject_to
raise RuntimeError(
RuntimeError: The problem is infeasible due to a constraint that always evaluates False.
This can happen if you've frozen too many decision variables, leading to an overconstrained problem.

The strange thing is that when remove the constraint and check it after the optimization is finished it is satisfied.

Any help is very much appreciated!

Thanks a lot, Enno

[peterdsharpe]

Thanks for the kind words!

Notice how the form of your constraint is fundamentally different than the rest of the constraints: a np.all() call forces symbolic variables (i.e., casadi.MX types) to be made concrete into static bool values - that's where this error is coming from.

To add this constraint, you'd want to compute Cp at each station using {upper, lower}_bl_ue/vinf_{i} + the outer-flow constituent relation (either Bernoulli or isentropic equations, depending on your case) and then set opti.subject_to(Cp_at_all_stations > -8).

[Roehrich]

Hi Peter, thanks for your help.
Maybe I still have not understood correctly but I was hoping that I can use the already calculated aero["Cpmin"] in the same way as aero[" CL"] or aero["analysis_confidence"] is used in the constraints. The only difference to the first case would be that I have to do the Boolean operation on an the Cpmin array with >= instead of ==.
Changed my constraints to
CP_min_constraints = np.array([-8, -8, -8, -8, -8, -2])
and the definition to
aero["Cpmin"][5] >= CP_min_constraints[5],
and it works. So the question is: should I add 6 of those definitions to capture all of my constraints or is there a "smarter" way to do this?

That brings me to a second question. the Cpmin value calculated by neuralfoil is -1.999 - spot on.

But in Xfoil the max is somewhat higher. It looks like that neuralfoil only calculates boundary layer values for 32 locations. It further looks like that this cannot be influenced as it is frozen into the training data. I tried to find the x-locations of those data points.
I found in neuralfoil main.py this:
These are the x-coordinates on the top and bottom surfaces of the airfoil where detailed boundary layer data is computed. It is made externally-accessible (nf.bl_x_points) in case you want to use it dynamically.
but I cannot find something equivalent in aerosandbox.
Can you help me to find it in my instances?
I guess I have to understand those values as an "average" over the region, right?

Thanks again for you help!

Enno