Add and compile new primal solver

[sodabryant]

@friedenhe Dear Dr.He, I tried to add a new primal solver named myDASimpleFoam, and here is my update process:

. $HOME/dafoam/loadDAFoam.sh
cd /home/lxl/dafoam/repos/dafoam-2.2.5/src/adjoint
./Allmake
cd $HOME/dafoam/repos/dafoam-2.2.5
./Allmake
source $HOME/dafoam/OpenFOAM/OpenFOAM-v1812-AD/etc/bashrc
./Allclean && ./Allmake 2> errorLog.txt
pip install .

The compilation last a real long time, then no error have been found. But, when I use the new myDASimpleFoam to run the pitzDaily case, I encountered the following error:

---

| DAFoam v2.2.5 |

Traceback (most recent call last):
File "runScript.py", line 153, in
DASolver = PYDAFOAM(options=daOptions, comm=gcomm)
File "/home/lxl/dafoam/packages/miniconda3/lib/python3.6/site-packages/dafoam/pyDAFoam.py", line 635, in init
self._initSolver()
File "/home/lxl/dafoam/packages/miniconda3/lib/python3.6/site-packages/dafoam/pyDAFoam.py", line 2162, in _initSolver
from .pyDASolverIncompressible import pyDASolvers
ImportError: dynamic module does not define module export function (PyInit_pyDASolverIncompressible)

I have no idea about this and don't know how to operate correctly. I am wondering whether the compile process is correct. Could you please offer me some assistance?

[friedenhe]

Can you elaborate on what exactly have you changed/added to add a new solver? It is likely that you missed some steps.

[sodabryant]

Yes sir! Except the existing solver, like DAsimpleFoam, DARhoSimpleFoam, etc. I wrote a new solver named myDASimpleFoam. Indeed, I want to try more operations on the "alphaPorosity" field, that's why I made a new solver.
And here gives my modification process:

1. in dafoam/repos/dafoam-2.2.5/src/adjoint/DAResidual, I add DAResidualMySimpleFoam.C and DAResidualMySimpleFoam.H.
2. in dafoam/repos/dafoam-2.2.5/src/adjoint/DASolver, I add myDASimpleFoam/MyDASimpleFoam.C and myDASimpleFoam/MyDASimpleFoam.H.
3. in dafoam/repos/dafoam-2.2.5/src/adjoint/DAStateInfo, I add DAStateInfoMySimpleFoam.C and DAStateInfoMySimpleFoam.H.
4. in dafoam/repos/dafoam-2.2.5/src/adjoint/Make/files_Incompressible, I add 3lines as follows:
   DASolver/myDASimpleFoam/myDASimpleFoam.C
   DAStateInfo/DAStateInfoMySimpleFoam.C
   DAResidual/DAResidualMySimpleFoam.C
5. in dafoam/repos/dafoam-2.2.5, compile it with the command: ./Allmake
6. source $HOME/dafoam/OpenFOAM/OpenFOAM-v1812-AD/etc/bashrc
   ./Allclean && ./Allmake 2> errorLog.txt
7. pip install .
   After the operations above, I modified the file runScript.py, then I got the error as showed above. So where's my mistake?

Hope for your reply. Thank you！

[friedenhe]

The overall steps look right. Did you change the class names in the new files? Did you change the TypeName(); in the added .H files accordingly?

[sodabryant]

Thanks for your reply, I tried three times and now it works, maybe I missed some details before.
Regards!

[sodabryant]

@friedenhe Dear Dr. He, I observed a complete optimization process of the pitzDaily case, but what bothers me now is that I can't find any criteria or conditions for ending up the optimization. Of course, the process automatively ends up when it fulfill the max iteration step 50 I defined. But it must have some other conditions to operate, Especially, what if I want to end up the optimization when the whole variable field, like alphaPorosity, not change anymore. Can you give me some hints about this?
Thank you!

[friedenhe]

Which optimizer did you use? You can definitely increase the max number of iterations to 50+ in the runScript.py.

[sodabryant]

I use IPOPT as the optimizer for now, but it shows a relatively worse result comparing with the result you presented before. Maybe it indeed needs more iterations? Here is the alphaPorosity field after 50 steps.

[friedenhe]

We only tested this case with SNOPT. We tried IPOPT and it did not work well for us. You may want to tweak the IPOPT parameters to see if they can improve the optimization.

[friedenhe]

BTW: you can find the optimality from IPOPT. Check the DAFoam workshop slides and video.

[sodabryant]

@friedenhe Dear He, I noticed that in your paper, you said that ADflow is a compressible solver, and DAFoam is a incompressible solver. But I found that the rhoSimpleFoam is included in the DAFoam, actually, I want to ask that if I want to further research the compressible flow with rhoCentralFoam, we know that rhoCentralFoam is a density based solver, is it possible to add rhoCentralFoam into the current DAFoam? Which state variables and residual equations should I define? Could you please offer me some assistance?

[friedenhe]

We added a few new features after the paper, and DAFoam supports subsonic and transonic cases now. But it does not work for supersonic flow yet. rhoCentralFoam is very different from the solvers already differentiated in DAFoam because it uses LTS. So we haven't gotten a chance to think about how to best differentiate it yet..