Enhancing Starting Point Selection in SU2 Inverse Regression

[VatshVan]

Could you explain the main challenges you’ve faced with the current starting point selection for inverse regression in SU2, particularly when dealing with nonlinear fluid behaviors near phase transitions?
I'm interested in understanding the specific issues that lead to unreliable predictions, and how you envision a data-driven, physics-informed neural network approach might enhance both the robustness and efficiency of the method.
Additionally, could you share insights on potential strategies for automating the selection process while balancing computational cost with prediction accuracy?