|
[Sponsors] |
March 10, 2015, 06:09 |
SU2 Shape optimization questions
|
#1 |
New Member
Adrien
Join Date: Mar 2015
Posts: 4
Rep Power: 11 |
Hi everyone,
I am currently working on SU2 in order to optimize the shape of an airfoil for a low-reynolds application. I have several questions regarding the setting of some variables and about the process of optimization. My case is to maximize the lift to drag ratio, while maintaing the lift coefficient above 1 and the thickness above 0.139. My config file is based on the test cases provided, except I changed the scale in order to have a objective function ~1e-5 and gradients norm ~1e-2. Is that scaling consistent for the optimizer? I also noticed that, at some point, SU2 was deforming the shape too much (the airfoil was no longer an airfoil) causing the solver to diverge. I changed BOUND_DV and DV_VALUE to 0.005. Is that the proper way to do it? I am not sure of what the DV_VALUE parameter is doing... I also have a question on the optimization procedure. From what I understood about the continous adjoint gradient method, the algorithm is basically: 1) solve the direct problem 2) solve the adjoint problem 3) compute the gradient 4) deform the shape according to the largest gradient and return to 1) In my case, I found that SU2 was performing these steps on the first iteration but then, for the following iterations it was only solving the direct problem, not the adjoint one. After a certain amount of iterations, it will evaluate the adjoint problem again. So my question is: why SU2 is not solving the adjoint problem and computing the gradient at each iteration? I attached my log file and my history_project.dat. As you can see, the efficiency is not really maximized. Is there something wrong with my settings? Thank you in advance, Adrien |
|
April 21, 2015, 11:44 |
|
#2 |
New Member
Daquan He
Join Date: Aug 2014
Posts: 7
Rep Power: 12 |
SU2 firstly calculate a gradient through adjoint method, then a line search is carried out in that direction. After several trials, an optimized point is found in this direction, and during the line search the gradient is not calculated. The iterations between 2 gradient evaluation steps are line search steps. This is my understanding of the algorithm.
|
|
April 21, 2015, 13:57 |
|
#3 |
New Member
Adrien
Join Date: Mar 2015
Posts: 4
Rep Power: 11 |
Hi,
Thanks for you answer. This seems to be working this way indeed. |
|
January 27, 2016, 01:51 |
i need su 2 shape optimizatin cfg file for su2 4.0.0
|
#4 |
New Member
balaji
Join Date: Jan 2016
Posts: 14
Rep Power: 10 |
i want do shape optimization by using python program please help how do and if i want to install python software
|
|
January 27, 2016, 05:02 |
|
#5 |
Super Moderator
Tim Albring
Join Date: Sep 2015
Posts: 195
Rep Power: 11 |
Hi bala732,
please consult the wiki first: https://github.com/su2code/SU2/wiki If there are more specific questions we can help you with that. |
|
January 27, 2016, 05:19 |
|
#6 |
New Member
balaji
Join Date: Jan 2016
Posts: 14
Rep Power: 10 |
thank u i found by python
|
|
January 27, 2016, 05:25 |
error in python coding
|
#7 |
New Member
balaji
Join Date: Jan 2016
Posts: 14
Rep Power: 10 |
this error are coming for my python how to rectify it my old_pr_ke4.jpgthis my codes and my error isold_pr_ke5.jpg
|
|
Tags |
adjoint, gradient, optimization, procedure, su2 |
|
|
Similar Threads | ||||
Thread | Thread Starter | Forum | Replies | Last Post |
shape deformation parametrization in SU2 | Mengmeng | SU2 Shape Design | 2 | October 22, 2014 21:56 |
Change the python file with shape optimization | 454514566@qq.com | SU2 | 1 | July 19, 2014 14:21 |
Airfoil Shape Optimization using gradient based methods | jaiswal | Main CFD Forum | 1 | April 10, 2014 12:22 |
Shape Optimization Error, Incorrect values produces by SU2_GDC | Dhig | SU2 Shape Design | 1 | March 20, 2014 06:58 |
Testcase for shape optimization using RANS simualtion | diwakaranant | SU2 Shape Design | 1 | April 18, 2013 14:27 |