Curious behavior with restart/solution files and adjoint

CarlosLozano · February 9, 2018, 07:47

Hi,
I've had some problems with the adjoint solver related to the restart file/ solution file scheme. When you want to get an adjoint solution, you first run the flow solver, which produces a restart_flow.dat file. This file has to be renamed to solution_flow.dat as input for SU2_SOL (to produce oputput files) and SU2_CFD to compute the adjoint. The problem arises when (for whatever reason) one does not rename the restart file and another solution_ file is present from a different case/mesh. The solver does not check for consistency and swallows up the wrong solution file producing a wrong or divergent adjoint solution.
regards,

fpalacios · February 24, 2018, 13:30

Agree, the renaming from restart to solution was done to prevent the accidental replacement of a good solution by a bad solution. 90% of the time we run the adjoint using the python scripts continuous_adjoint.py or discrete_adjoint.py these scripts rename the files in a consistent way. However some people like to just use the same name for
SOLUTION_FLOW_FILENAME= solution_flow.dat
SOLUTION_ADJ_FILENAME= solution_adj.dat
RESTART_FLOW_FILENAME= restart_flow.dat
RESTART_ADJ_FILENAME= restart_adj.dat

so the described case is not possible.

Best,
Francisco

By the way, if you are thinking developing the software we have recently created the SU2 IDS www.su2devsociety.org that could be interesting for you.

February 9, 2018, 07:47	Curious behavior with restart/solution files and adjoint	#1
CarlosLozano New Member Join Date: Feb 2018 Posts: 27 Rep Power: 8	Hi, I've had some problems with the adjoint solver related to the restart file/ solution file scheme. When you want to get an adjoint solution, you first run the flow solver, which produces a restart_flow.dat file. This file has to be renamed to solution_flow.dat as input for SU2_SOL (to produce oputput files) and SU2_CFD to compute the adjoint. The problem arises when (for whatever reason) one does not rename the restart file and another solution_ file is present from a different case/mesh. The solver does not check for consistency and swallows up the wrong solution file producing a wrong or divergent adjoint solution. regards,

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
Problems launching adjoint	max_	SU2	2	March 12, 2016 12:32
Unsteady simulation solution files in parallel	gunnersnroses	SU2	1	December 15, 2015 14:28
hard to convergent with SU2 4.0 for Euler adjoint	Xianguu	SU2	1	July 14, 2015 18:09

February 24, 2018, 13:30		#2
fpalacios Super Moderator Francisco Palacios Join Date: Jan 2013 Location: Long Beach, CA Posts: 404 Rep Power: 15	Agree, the renaming from restart to solution was done to prevent the accidental replacement of a good solution by a bad solution. 90% of the time we run the adjoint using the python scripts continuous_adjoint.py or discrete_adjoint.py these scripts rename the files in a consistent way. However some people like to just use the same name for SOLUTION_FLOW_FILENAME= solution_flow.dat SOLUTION_ADJ_FILENAME= solution_adj.dat RESTART_FLOW_FILENAME= restart_flow.dat RESTART_ADJ_FILENAME= restart_adj.dat so the described case is not possible. Best, Francisco By the way, if you are thinking developing the software we have recently created the SU2 IDS www.su2devsociety.org that could be interesting for you.