[RobVM]

Dear SU2 community,


I am currently using the SU2 version 6.0.0. for an airfoil optimization problem using the discrete adjoint method. Due to the nature of my mesh, I am having problems converging the solution for the DIRECT problem using Multigrid, but in the ADJOINT the Multigrid is really helpful.



I saw in the config_template.cfg that you can turn on or off the Multigrid option in the Adjoint problem. (MG_ADJFLOW = YES).



Is there any way to turn off the Multigrid option for the DIRECT Problem and use it just in the ADJOINT solution by imposing MG_ADJFLOW = YES?


Thank you.

[pcg]

What MG settings are you using and what is the size/type of your mesh? Maybe with some tuning you would be able to use it for both problems.

[RobVM]

I finally skipped the use of the Multigrid. It demands a bit more of computational time without it but the convergence is satisfactory.
The problems that I have now are related to the optimization itself. Even if I set the airfoil thickness constraint to be greater than 0.05 (5%), in the 10th iteration the upper and lower surfaces cross with each other in the leading edge, hence leading to divergence.



I'm using FFD_CONTROL_POINT_2D Design variables, as follows in the Optimal shape optimization:


% --------------------- OPTIMAL SHAPE DESIGN DEFINITION -----------------------%
% Available Objective functions
% DRAG, LIFT, SIDEFORCE, PRESSURE, FORCE_X, FORCE_Y,
% FORCE_Z, MOMENT_X, MOMENT_Y, MOMENT_Z, EFFICIENCY,
% EQUIVALENT_AREA, THRUST, TORQUE, FREESURFACE
% Optimization objective function with scaling factor, separated by semicolons.
% To include quadratic penalty function: use OPT_CONSTRAINT option syntax within the OPT_OBJECTIVE list.
% ex= Objective * Scale
OPT_OBJECTIVE= DRAG*0.001
%
% Optimization constraint functions with pushing factors (affects its value, not the gradient in the python scripts), separated by semicolons
% ex= (Objective = Value ) * Scale, use '>','<','='
OPT_CONSTRAINT= (LIFT > 2.865440)*0.001; (AIRFOIL_THICKNESS > 0.05)*0.001
%
% Factor to reduce the norm of the gradient (affects the objective function and gradient in the python scripts)
% In general, a norm of the gradient ~1E-6 is desired.
%OPT_GRADIENT_FACTOR= 1E-6
%
% Factor to relax or accelerate the optimizer convergence (affects the line search in SU2_DEF)
% In general, surface deformations of 0.01'' or 0.0001m are desirable
%OPT_RELAX_FACTOR= 1E2
%
% Maximum number of optimizer iterations ( 100 by default)
OPT_ITERATIONS= 30
%
% Requested accuracy ( 1e-10 by default)
OPT_ACCURACY= 1E-6
%
% Upper bound for each design variable (bound in the python optimizer)
OPT_BOUND_UPPER= 1E6
%
% Lower bound for each design variable (bound in the python optimizer)
OPT_BOUND_LOWER= -1E6
%
% Optimization design variables, separated by semicolons
DEFINITION_DV= ( 15, 1.0 | AIRFOIL | AIRFOIL_BOX,0,0,0.0,15.0);( 15, 1.0 | AIRFOIL | AIRFOIL_BOX,1,0,0.0,15.0);( 15, 1.0 | AIRFOIL | AIRFOIL_BOX,2,0,0.0,15.0);( 15, 1.0 | AIRFOIL | AIRFOIL_BOX,3,0,0.0,15.0);( 15, 1.0 | AIRFOIL | AIRFOIL_BOX,4,0,0.0,15.0);( 15, 1.0 | AIRFOIL | AIRFOIL_BOX,5,0,0.0,15.0);( 15, 1.0 | AIRFOIL | AIRFOIL_BOX,6,0,0.0,15.0);( 15, 1.0 | AIRFOIL | AIRFOIL_BOX,7,0,0.0,15.0);( 15, 1.0 | AIRFOIL | AIRFOIL_BOX,8,0,0.0,15.0);( 15, 1.0 | AIRFOIL | AIRFOIL_BOX,9,0,0.0,15.0);( 15, 1.0 | AIRFOIL | AIRFOIL_BOX,10,0,0.0,15.0);( 15, 1.0 | AIRFOIL | AIRFOIL_BOX,0,1,0.0,15.0);( 15, 1.0 | AIRFOIL | AIRFOIL_BOX,1,1,0.0,15.0);( 15, 1.0 | AIRFOIL | AIRFOIL_BOX,2,1,0.0,15.0);( 15, 1.0 | AIRFOIL | AIRFOIL_BOX,3,1,0.0,15.0);( 15, 1.0 | AIRFOIL | AIRFOIL_BOX,4,1,0.0,15.0);( 15, 1.0 | AIRFOIL | AIRFOIL_BOX,5,1,0.0,15.0);( 15, 1.0 | AIRFOIL | AIRFOIL_BOX,6,1,0.0,15.0);( 15, 1.0 | AIRFOIL | AIRFOIL_BOX,7,1,0.0,15.0);( 15, 1.0 | AIRFOIL | AIRFOIL_BOX,8,1,0.0,15.0);( 15, 1.0 | AIRFOIL | AIRFOIL_BOX,9,1,0.0,15.0);( 15, 1.0 | AIRFOIL | AIRFOIL_BOX,10,1,0.0,15.0);



In order to let a wide range of movement freedom to the DV, I set the OPT_BOUNDS to a high number (+-1E6). Moreover, in the x_Mov, y_Mov of the DV, I set the movement on y (the only direction allowed to move) to 15.0, in order to let it explore a wider range of possibilities.


The problem is that the surfaces cross with each other, violating the thickness constraint.

Maybe the freedom to the DV should be a bit more restrictive and hence use a lower number, but I don't know how this OPT_BOUND and x_Mov,y_Mov variables work at all.


1. Shouldn't the thickness constraint be respected even if the allowed movement is wide?
2. Are the OPT_BOUNDS the physical limit of the maximum deformation of the DV on the FFD box? In which units are they expressed?
3. In which units are the x_Mov and y_Mov represented, in terms of chords? (i.e. if y_Mov=1, that means the allowed vertical movement of that certain DV is one chord? or they're the allowed maximum displacement per Opt iteration?)


Let's see if I can get some help in these issues.


Thank you and sorry for my late reply.