Discrete adjoint gradients with custom parameterization

asthelen · July 13, 2020, 13:09

Hello,

My geometry is parameterized externally, and I have the corresponding grid velocities on the surface mesh, but I can't get the discrete adjoint solver to compute gradients using these external grid velocities. I set DV_KIND='SURFACE_FILE' and DV_SENS_FILENAME equal to the grid velocity filename (surface coordinates and x,y,z displacements corresponding to each DV), but that didn't seem to do what I want, as it always defaults to one Hicks Henne bump.

Is what I'm trying to do possible in SU2 6.2.0 (or newer)? If so, what am I doing wrong?

Thanks.

Scramjetdk · July 19, 2020, 11:15

Hi, Andrew,

Form my knowledge, if you are using an external shape parameterization method, and want to deform the mesh by using SU2_DEF, you can set as following parameters:

DV_KIND=SURFACE_FILE
DV_MARKER=( airfoil )
DV_PARAM=( 1.0 )
DV_FILENAME= surface_file.dat

wherein the surface_file.dat, the format is
GlobalID_0, x_0, y_0, z_0
GlobalID_1, x_1, y_1, z_1
...
GlobalID_N, x_N, y_N, z_N
where N is the total number of vertices on all moving markers, and x/y/z are
the new position of each vertex. Points can be in any order.

However, I am sorry about, I have no idea how to set DV_SENS_FILENAME.

At last, please remember to delete all the keywords about 'OPTIMAL SHAPE DESIGN DEFINITION'. As you are not asking the python optimization algorithm finding the optimal results.

Hope the given information is helpful.
Best

July 13, 2020, 13:09	Discrete adjoint gradients with custom parameterization	#1
asthelen New Member Andrew Join Date: Mar 2015 Posts: 5 Rep Power: 11	Hello, My geometry is parameterized externally, and I have the corresponding grid velocities on the surface mesh, but I can't get the discrete adjoint solver to compute gradients using these external grid velocities. I set DV_KIND='SURFACE_FILE' and DV_SENS_FILENAME equal to the grid velocity filename (surface coordinates and x,y,z displacements corresponding to each DV), but that didn't seem to do what I want, as it always defaults to one Hicks Henne bump. Is what I'm trying to do possible in SU2 6.2.0 (or newer)? If so, what am I doing wrong? Thanks.

Similar Threads
Thread	Thread Starter	Forum	Replies	Last Post
the gradients calculated through discrete adjoint are near zero	Haowang	SU2	9	October 17, 2019 23:53
RANS discrete adjoint oscillating	pgdelaserna	SU2	3	August 25, 2019 05:32
Discrete vs Continuous Adjoint solution variable: Scaling/normalization issue?	asitav	SU2	2	November 21, 2018 20:11
Discrete Adjoint giving garbage values	Manish Kumar Mishra	SU2	1	March 7, 2018 16:35
Shape optimization doesn't converge - Adjoint log file	JPBLourenco	SU2 Shape Design	0	December 11, 2017 09:41

July 19, 2020, 11:15		#2
Scramjetdk New Member Zhenkai Zhang Join Date: Oct 2019 Posts: 3 Rep Power: 7	Hi, Andrew, Form my knowledge, if you are using an external shape parameterization method, and want to deform the mesh by using SU2_DEF, you can set as following parameters: DV_KIND=SURFACE_FILE DV_MARKER=( airfoil ) DV_PARAM=( 1.0 ) DV_FILENAME= surface_file.dat wherein the surface_file.dat, the format is GlobalID_0, x_0, y_0, z_0 GlobalID_1, x_1, y_1, z_1 ... GlobalID_N, x_N, y_N, z_N where N is the total number of vertices on all moving markers, and x/y/z are the new position of each vertex. Points can be in any order. However, I am sorry about, I have no idea how to set DV_SENS_FILENAME. At last, please remember to delete all the keywords about 'OPTIMAL SHAPE DESIGN DEFINITION'. As you are not asking the python optimization algorithm finding the optimal results. Hope the given information is helpful. Best