Good afternoon! Thank you for your responses.
Speaking in more detail about the idea that I wanted to implement, it was supposed to be a plastic TIR element that works in a non-imaging lighting system. The global goal was to create an optical system with a variable angle of radiation divergence and an approximately uniform distribution of illumination. I was wondering if I could use the input surface, the total internal reflection surface and the output surface as optimization surfaces by defining them in the CAD model as surfaces of revolution, the profile of which is given by the equation. Then the coefficients of the equation could be used as optimization parameters.

The figure shows a sketch of the model, the optimization of which I planned to carry out. This is not a working option because the sketch curves are fixed. It was assumed that the near axial part is calculated in sequential mode and defined in CAD as well as the TIR surface using the surface equation. I think I wanted a too simple solution, since I thought about such a trivial solution as optimizing a single CAD model.
At the moment, there is an understanding that perhaps a combination of inconsistent "Even Asphere", "Standard lens" and "Freeform Z" surfaces can serve as a solution in such a way that each surface will roughly recreate a part of the CAD component.
I'm still learning the program and I like that there are different ways to solve problems.

Hi Devorse,

I had a quick chat with one of my colleagues and we think it is a limitation of the CAD link. That being said, my colleague will check next week as he is our CAD expert.

I was wondering if you could define your CAD part with an OpticStudio built-in model instead. For example we have the Freeform object where you define a profile (sag points) along YZ and then the software does a rotation. 

Would you have a couple of screenshots to share to give us some more information? Thank you.