[Raj_1]

Hi,
Recently I started exploring possibility of using SU2 for robust optimization where I need to run the direct and adjoint solver of SU2 at multiple sampling points at each iteration of optimization process and need to feed a modified gradient of the objective to the optimizer.

I guess I need to write few scripts to do it. If so, it would be great if someone can provide me some ideas to implement it as:

1. How can I call SU2_CFD multiple times for each iteration of the optimization process.
2. How to modify objective gradient wrt design var. in the iteration process.

Thanks
Raj

[fpalacios]

Hi Raj,

What you are proposing is a deep modification of the optimization algorithm. There are different ways to proceed and, obviously, the first step is to be familiar with the python part of the code. I think it would be interesting if you take a look at SU2_PY/SU2/opt/scipy_tools.py and SU2_PY/compute_polar.py

The first file is the highest level to define an optimization strategy... in this case SLSQP using Scipy. The second file is just an example of how to run a polar.

Best,
Francisco Palacios

[Raj_1]

Dear Francisco,

Thank you very much for you kind suggestions. I will try all possibilities.

Regards,
Dinesh

[Arad]

Hi Raj,

I find your question very intersting and important. Especially if one looks at multi-points
optimization and and an objective function/restraints that are evaluated at different flow conditions. Would you like to share your experience in this field ?

Best regards,

Eran