SU2 Discrete adjoint skips optimization iterations

StanislavKarpuk · November 17, 2019, 14:18

I have been trying to perform an unconstrained aerodynamic optimization of a BWB using SU2 Descrete Adjoint Optimization. I have noticed that the adjoint was calculated for the first 3 iterations, then it was not calculated for the 4th one, then it was solved again and then stopped being solved completely. Could you tell me what could cause such behavior and if it is normal.

Thank you,
Stan

atelcikti1 · November 18, 2019, 07:26

Yes, it is possible. When adjoint solver is called this means optimizer needs gradients of design variables, when it is not called optimizer simply do line search to decide new stepsize. in this case only flow solution is enough.

StanislavKarpuk · November 18, 2019, 07:54

atelcikti1,
Thank you very much! it make sense now

StanislavKarpuk · November 27, 2019, 05:44

I have one more question. At some point the optimizer stops making adjoint iterations and the drag stops reducing. However, the solution keeps going. Does it mean I am converged because I see some deformations of the geometry but no effect on drag?

November 17, 2019, 14:18	SU2 Discrete adjoint skips optimization iterations	#1
StanislavKarpuk New Member Stanislav Karpuk Join Date: Sep 2016 Posts: 9 Rep Power: 10	I have been trying to perform an unconstrained aerodynamic optimization of a BWB using SU2 Descrete Adjoint Optimization. I have noticed that the adjoint was calculated for the first 3 iterations, then it was not calculated for the 4th one, then it was solved again and then stopped being solved completely. Could you tell me what could cause such behavior and if it is normal. Thank you, Stan

Thread Tools	Search this Thread
Show Printable Version Email this Page	Search this Thread: Advanced Search
Display Modes
Linear Mode Switch to Hybrid Mode Switch to Threaded Mode

Similar Threads
Thread	Thread Starter	Forum	Replies	Last Post
laplacianFoam with source term	Herwig	OpenFOAM Running, Solving & CFD	17	November 19, 2019 13:47
Segmentation fault when using reactingFOAM for Fluids	Tommy Floessner	OpenFOAM Running, Solving & CFD	4	April 22, 2018 12:30
Free surface issues with interDyMFoam for hydroturbine	oumnion	OpenFOAM Running, Solving & CFD	0	October 6, 2017 14:05
Extrusion with OpenFoam problem No. Iterations 0	Lord Kelvin	OpenFOAM Running, Solving & CFD	8	March 28, 2016 11:08
Micro Scale Pore, icoFoam	gooya_kabir	OpenFOAM Running, Solving & CFD	2	November 2, 2013 13:58

November 18, 2019, 07:26		#2
atelcikti1 Member Join Date: Mar 2017 Posts: 61 Rep Power: 10	Yes, it is possible. When adjoint solver is called this means optimizer needs gradients of design variables, when it is not called optimizer simply do line search to decide new stepsize. in this case only flow solution is enough.

November 18, 2019, 07:54		#3
StanislavKarpuk New Member Stanislav Karpuk Join Date: Sep 2016 Posts: 9 Rep Power: 10	atelcikti1, Thank you very much! it make sense now

November 27, 2019, 05:44		#4
StanislavKarpuk New Member Stanislav Karpuk Join Date: Sep 2016 Posts: 9 Rep Power: 10	I have one more question. At some point the optimizer stops making adjoint iterations and the drag stops reducing. However, the solution keeps going. Does it mean I am converged because I see some deformations of the geometry but no effect on drag?