Shape optimization parameters

[giusam]

Hi, everyone
I am currently configuring an Aerodynamic Shape Optimization (ASO) case for an RAE2822 airfoil in transonic conditions (Ma=0.734,Re=6.5⋅106) using Hicks-Henne functions and discrete adjoint approach.

I have specific questions regarding the following settings in the configuration file:

1. Function of Scaling Factors

Could you explain exactly how the multipliers in OPT_CONSTRAINT (e.g., * 0.001) affect the optimizer's internal logic?

Observations: When using low scaling factors, I achieved a reduction in DRAG, but the AIRFOIL_AREA constraint was violated by approximately 4%.

Issue: When I attempt to enforce the constraints more strictly (by increasing the scaling factors to 1), the optimization tends to stall after a few iteration with a GNORM that remains unchanged.

Question: Do these factors only scale the value, or do they influence the weighting relative to the objective function in the Lagrangian?

2. Gradient and Relaxation Parameters

   OPT_GRADIENT_FACTOR (1E-6): From what I understand, this factor is used to ensure that the initial GNORM is on the order of 10−6. Is this correct?

   OPT_RELAX_FACTOR (1E2) and OPT_LINE_SEARCH_BOUND (1E6): I believe these parameters control the surface deformation process. Could you clarify their specific roles?

   OPT_BOUND_UPPER (1E10) and OPT_BOUND_LOWER (-1E10): I am unclear on what these specific bounds regulate within the optimization loop.

3. Setting Rules and Best Practices

Is there a recommended "rule of thumb" or a specific procedure that you use to determine the optimal values for:

Scaling factors?

OPT_LINE_SEARCH_BOUND and OPT_RELAX_FACTOR?

OPT_BOUND_UPPER/LOWER?

Goal: How should I adjust these terms to ensure strict adherence to the constraints (avoiding the 4% violation) without causing the optimizer to stall?
Thanks for your kind attention and I hope you have a nice day,
Giuseppe.