Adjoint Calculation Error

tomp1993 · August 23, 2017, 12:55

Hi,
I am running a 2D aerofoil optimization using V5.0.0;
the flow solution is converged but each log_adjoint.out file shows the following:

Code:

log10[Maximum residual]: -inf.
Maximum residual point 0, located at (0, 0).

 Iter    Time(s)   Res[Psi_Rho]     Res[Psi_E]      Sens_Geo      Sens_AoA
    0   0.524635     -32.000000     -32.000000    0.0000e+00    0.0000e+00
CSysSolve::FGMRES(): system solved by initial guess.

when in fact the flow solution file shows the following:

Code:

----------------------- Residual Evolution Summary ----------------------
log10[Maximum residual]: -2.83626.
Maximum residual point 11866, located at (0.191485, -0.0191805).
----------------------------------------------------------------------

just before writing and exiting.
the adjoint gradients are then calculated to be zero (i think as a result of the above issue)
any help would be highly appreciated!
if more information on the .cfg file or .su2 grid is needed let me know.
Many thanks!
Tom

AQuiros · August 24, 2017, 18:45

What are the values you are using in the reduction factors of the CFL? This issue can happen if they are 0 since you'll be using a CFL of 0.

tomp1993 · August 29, 2017, 13:22

Hi,
the CFL was positive with a value of around 0.5-1.5.
the reduction factors were around 1.0-0.9.
I have now gotten around the problem by modifying a cfg file from the NACA0012 steady optimization Testcase for use with my own grid and conditions.
However, i am now facing the problem of the design returning to its original condition after a single 'good' design iteration. This issue is being discussed in another Thread.
Thanks for the assistance.
Tom

August 23, 2017, 12:55	Adjoint Calculation Error	#1
tomp1993 Member tom pratt Join Date: Jan 2016 Posts: 55 Rep Power: 10	Hi, I am running a 2D aerofoil optimization using V5.0.0; the flow solution is converged but each log_adjoint.out file shows the following: Code: log10[Maximum residual]: -inf. Maximum residual point 0, located at (0, 0). Iter Time(s) Res[Psi_Rho] Res[Psi_E] Sens_Geo Sens_AoA 0 0.524635 -32.000000 -32.000000 0.0000e+00 0.0000e+00 CSysSolve::FGMRES(): system solved by initial guess. when in fact the flow solution file shows the following: Code: ----------------------- Residual Evolution Summary ---------------------- log10[Maximum residual]: -2.83626. Maximum residual point 11866, located at (0.191485, -0.0191805). ---------------------------------------------------------------------- just before writing and exiting. the adjoint gradients are then calculated to be zero (i think as a result of the above issue) any help would be highly appreciated! if more information on the .cfg file or .su2 grid is needed let me know. Many thanks! Tom

Similar Threads
Thread	Thread Starter	Forum	Replies	Last Post
Compile calcMassFlowC	aurore	OpenFOAM Programming & Development	13	March 23, 2018 08:43
DPM udf error	haghshenasfard	FLUENT	0	April 13, 2016 07:35
Undeclared Identifier Errof UDF	SteveGoat	Fluent UDF and Scheme Programming	7	October 15, 2014 08:11
Compile problem	ivanyao	OpenFOAM Running, Solving & CFD	1	October 12, 2012 10:31
How to get the max value of the whole field	waynezw0618	OpenFOAM Running, Solving & CFD	4	June 17, 2008 06:07

August 24, 2017, 18:45		#2
AQuiros New Member Alejandro Quirós Join Date: Nov 2016 Location: Chiclana, Spain Posts: 6 Rep Power: 9	What are the values you are using in the reduction factors of the CFL? This issue can happen if they are 0 since you'll be using a CFL of 0.

August 29, 2017, 13:22		#3
tomp1993 Member tom pratt Join Date: Jan 2016 Posts: 55 Rep Power: 10	Hi, the CFL was positive with a value of around 0.5-1.5. the reduction factors were around 1.0-0.9. I have now gotten around the problem by modifying a cfg file from the NACA0012 steady optimization Testcase for use with my own grid and conditions. However, i am now facing the problem of the design returning to its original condition after a single 'good' design iteration. This issue is being discussed in another Thread. Thanks for the assistance. Tom