
[Sponsors] 
July 18, 2011, 15:02 
adjointShapeOptimizationFoam

#1 
New Member
SamboReal
Join Date: Feb 2011
Posts: 9
Rep Power: 7 
Hello guys,
I have a questions depending on adjointShapeOptimizationFoam. Did anyone of you already used it?. I runned the tutorial but I can not identify that he is improving the geometry. I didn’t changed anything in the tutorial! Can someone help me? Thanks 

July 18, 2011, 16:42 

#2 
New Member
SamboReal
Join Date: Feb 2011
Posts: 9
Rep Power: 7 
?????????????


July 19, 2011, 00:32 

#3 
Member
Masashi Ohbuchi
Join Date: Oct 2009
Posts: 72
Rep Power: 8 
Hi,
You can plot contour of alpha. I suppose this means plugged space for drag reduction. And following presentation gives adjoint method for optimization. http://www.openfoamworkshop.org/6th_...ers_slides.pdf 

July 20, 2011, 09:45 

#4 
New Member
SamboReal
Join Date: Feb 2011
Posts: 9
Rep Power: 7 
Thank you very much.
Do you may no, where I can find some more information about that? 

August 25, 2011, 09:55 

#5 
Member
Nick Gardiner
Join Date: Apr 2009
Location: Chichester, UK
Posts: 93
Rep Power: 8 
Hi
Can you explain what the alpha stands for? And in the slide show linked to above, is L the same as lambda in the solver? Thanks 

October 20, 2011, 14:15 

#6 
Senior Member
David Boger
Join Date: Mar 2009
Location: Penn State Applied Research Laboratory
Posts: 146
Rep Power: 8 
Hi Nick,
You've probably answered your own questions long ago (or given up entirely), but no, the L in the slides refers to the cost function that is to be optimized. The purpose of the equations is to calculate the gradient of the cost function with respect to the porosity (alpha) in each cell. Once the gradient is in hand, the alpha field is updated by moving in the direction of the gradient (steepest descent). The lambda in the code controls the step size taken along the gradient; i.e., the step size for the steepest descent method. Hope that helps, David
__________________
David A. Boger 

September 10, 2012, 20:29 

#7 
New Member
Erkan
Join Date: May 2010
Posts: 9
Rep Power: 7 
Hi guys,
i am also new with openfoam. Unfortunatelly the above links are not working . Is there any other document where too have a look on the equations? Checking the .c file and trying to understand the equations is very difficult. thanks Last edited by uwsa; September 10, 2012 at 23:15. 

October 24, 2012, 12:06 

#8 
New Member
Join Date: Mar 2009
Location: Germany
Posts: 17
Rep Power: 8 
Hello to everybody,
has someone already tried to solve shape optimization for an alternative cost function like flow uniformity? As far as I understand, the current implementation tries to reduce total pressure loss. Has someone a starting point for me to change this or add something? Thanks in advance, georg 

October 25, 2012, 04:12 

#9 
Member
Roland
Join Date: Mar 2009
Location: Netherlands
Posts: 64
Rep Power: 8 
Hi Georg,
Some references which present additional cost functions: Othmer, C., Th. Kaminski, and R. Giering, Computation of topological sensitivities in fluid dynamics: Cost function versatility, Proc. ECCOMAS CFD, Delft, 2006 http://www.wire.tubs.de/mitarbeiter...ad_eccomas.pdf and Othmer, C. (2008), A continuous adjoint formulation for the computation of topological and surface sensitivities of ducted flows. International Journal for Numerical Methods in Fluids, 58: 861–877. doi: 10.1002/fld.1770 http://onlinelibrary.wiley.com/doi/1....1770/abstract regards, Sylvester 

October 26, 2012, 06:16 

#10 
New Member
Join Date: Mar 2009
Location: Germany
Posts: 17
Rep Power: 8 
Hi Sylvester,
thank you for the references. I did not know the first one yet. Some more stuff to read! Honestly, I have some difficulties to identify the definition of the cost function within the solver. Can you help? Best regards, georg 

October 26, 2012, 08:24 

#11 
Member
Roland
Join Date: Mar 2009
Location: Netherlands
Posts: 64
Rep Power: 8 
Hi Georg,
That is because the cost function is not defined in the solver, but in the boundary conditions. In this case they are 'adjointOutletPressure' and 'adjointOutletVelocity'. regards, Sylvester 

October 26, 2012, 11:12 

#12 
New Member
Join Date: Mar 2009
Location: Germany
Posts: 17
Rep Power: 8 
Sylvester, you´re right!
I found the explanation within the papers. Thanks for your pointer! Unfortunately, this does not solve my problem completely. Maybe I just do not see the wood for the trees. I am still puzzling around the translation of the cost function, e.g. pressure loss, into the correct boundary conditions! Any hint is helpful! Best regards, Georg 

Thread Tools  
Display Modes  

