|
[Sponsors] |
March 3, 2018, 08:25 |
Discrete Adjoint giving garbage values
|
#1 |
New Member
Manish Kumar Mishra
Join Date: May 2015
Posts: 1
Rep Power: 0 |
I was previously using continuous adjoint solver for gradient computation. I decided to try Discrete Adjoint once but I am facing a problem here.
My direct simulation works, so I have a flow solution for the adjoint solver to use. I am also able to use the Continuous Adjoint solver to compute the gradient but when I tried running the Discrete Adjoint solver, it runs and residual values drop by 10 orders but the solution is way far off and when I check the adjoint variable contour, it seems terrible. With discrete Adj solver, the solution blows up! I thought the problem was with my problem but I am facing the same problem with the config file and grid in QuickStart tutorial airfoil problem. Continuous adj works but not discrete adj. I tried changing a few setting but I think those shouldn't make any difference. Has anyone faced this problem? Or if you can guess it and suggest a few things to try, I will be glad. Thank you |
|
March 7, 2018, 16:35 |
|
#2 |
Member
Ole Burghardt
Join Date: Mar 2016
Location: Kiel, Germany
Posts: 60
Rep Power: 10 |
Assuming that your SU2_CFD_AD is fine and that you are reading in the solution file correctly - does changing the objective function change the behaviour?
Sounds a bit like a problem in that direction... |
|
Tags |
discrete adjoint, quickstart, tutorial |
|
|
Similar Threads | ||||
Thread | Thread Starter | Forum | Replies | Last Post |
Shape optimization doesn't converge - Adjoint log file | JPBLourenco | SU2 Shape Design | 0 | December 11, 2017 09:41 |
Discrete Values for Mesh Motion | UnderwaterAlex | CFX | 2 | February 21, 2008 19:21 |
Minimization discrete values. | nico | Main CFD Forum | 0 | December 14, 2006 12:03 |
Discrete Boundary Values | Rhydar | CFX | 0 | July 13, 2002 09:20 |
IPALL giving NULL values | cfd guy | CFX | 0 | November 29, 2001 11:51 |