CFD Online Logo CFD Online URL
www.cfd-online.com
[Sponsors]
Home > Forums > SU2

Sensetivities

Register Blogs Members List Search Today's Posts Mark Forums Read

Reply
 
LinkBack Thread Tools Display Modes
Old   October 2, 2016, 07:56
Default Sensetivities
  #1
Member
 
Tom
Join Date: Oct 2015
Posts: 67
Rep Power: 4
mlooti is on a distinguished road
Hello my friends

I’ve been involved with this problem for about a week and I’m still struggling to find a way around it
I have a wing which is look like the attached picture. I want to do some optimization on that based on let’s say drag objective function. I solved the RANS eq. on the geometry and the convergence went well. The residuals are:


Iter Time(s) Res[Rho] Res[nu] CLift(Total) CDrag(Total)
20000 2.646225 -7.977799 -10.375939 0.414779 0.296870

After that I solve the continuous adjoint equations (drag OF), which the residuals are:


Iter Time(s) Res[Psi_Rho] Res[Psi_E] Sens_Geo Sens_Mach
20000 4.535583 -3.265719 -9.096933 -3.0207e+03 -2.9336e-01

I investigate a lot of effort on the meshing part. The grid was generated using pointwise software. In this geometry there are not any noticeable sharp edges but I try to resolve any feature on the geometry (added pictures). As it is shown in the residuals the surface sensitivities are really high which lead to high gradients and high surface deformation.

can any body give a hint on that?

** in the countour the values are mostly around 30000
Attached Images
File Type: jpg 1.jpg (77.0 KB, 10 views)
File Type: jpg 24.jpg (40.3 KB, 12 views)
mlooti is offline   Reply With Quote

Old   March 11, 2017, 04:36
Default
  #2
hlk
Senior Member
 
Heather Kline
Join Date: Jun 2013
Posts: 267
Rep Power: 7
hlk is on a distinguished road
Quote:
Originally Posted by mlooti View Post
Hello my friends

Iíve been involved with this problem for about a week and Iím still struggling to find a way around it
I have a wing which is look like the attached picture. I want to do some optimization on that based on letís say drag objective function. I solved the RANS eq. on the geometry and the convergence went well. The residuals are:


Iter Time(s) Res[Rho] Res[nu] CLift(Total) CDrag(Total)
20000 2.646225 -7.977799 -10.375939 0.414779 0.296870

After that I solve the continuous adjoint equations (drag OF), which the residuals are:


Iter Time(s) Res[Psi_Rho] Res[Psi_E] Sens_Geo Sens_Mach
20000 4.535583 -3.265719 -9.096933 -3.0207e+03 -2.9336e-01

I investigate a lot of effort on the meshing part. The grid was generated using pointwise software. In this geometry there are not any noticeable sharp edges but I try to resolve any feature on the geometry (added pictures). As it is shown in the residuals the surface sensitivities are really high which lead to high gradients and high surface deformation.

can any body give a hint on that?

** in the countour the values are mostly around 30000
Using the continuous adjoint, the following settings can be useful:
CFL_REDUCTION_ADJFLOW = (something less than 1 to use a smaller CFL than the direct problem)
LIMIT_ADJFLOW= (default 1E6, can test out other values)
MG_ADJFLOW = YES by default, however NO can be useful when convergence is difficult
RELAXATION_FACTOR_ADJFLOW = something slightly less than one if there is an issue with the mesh
SPATIAL_ORDER_ADJFLOW = 1ST_ORDER often has an easier time converging.
Although it is recommended to use the same numerical method setting for direct and adjoint, you can also play around with that if your problem is not converging well.
hlk is offline   Reply With Quote

Reply

Thread Tools
Display Modes

Posting Rules
You may not post new threads
You may not post replies
You may not post attachments
You may not edit your posts

BB code is On
Smilies are On
[IMG] code is On
HTML code is Off
Trackbacks are On
Pingbacks are On
Refbacks are On



All times are GMT -4. The time now is 05:10.