Error writing out the sensitivity map

[uttamcadambi07]

Hi,

I am using DAFoam v3.0.6 and I am interested in writing out the sensitivity map.

I know that the v3 script does not allow for this so I use the v2 script.

I tested the case for the field inversion for a NACA0018 airfoil (i did runAdjoint) and got a file named totalDerivHist.txt

But now, I tried the same for fvSource as a design variable but there seems to be some issue in the function that is being called in pyDAFoam.py that writes out the derivatives

here is the error

+--------------------------------------------------------------------------+
|              Evaluating Objective Function Sensitivities 001             |
+--------------------------------------------------------------------------+
writeDeformedFFDs is deprecated since v3.0.1!
Moving time 316 to 0.00000001
Running adjoint Solver 001
dRdWT Jacobian Free created!
Computing dRdWTPC 509 s
Initializing dRdWCon. 509 s
dRdWCon Created. 552 s
Reading Coloring dRdWColoring_1
Validating Coloring...
No Conflicting Colors Found!
Partial derivative matrix created. 552 s
dRdWTPC: 0 of 353, ExecutionTime: 553 s
dRdWTPC: 100 of 353, ExecutionTime: 601 s
dRdWTPC: 200 of 353, ExecutionTime: 653 s
dRdWTPC: 300 of 353, ExecutionTime: 702 s
dRdWTPC: 352 of 353, ExecutionTime: 727 s
Solver Type: gmres
GMRES Restart: 1000
ASM Overlap: 1
Global PC Iters: 0
Local PC Iters: 1
Mat ReOrdering: natural
ILU PC Fill Level: 2
GMRES Max Iterations: 1000
GMRES Relative Tolerance: 1e-10
GMRES Absolute Tolerance: 1e-15
Calculating dFdW using reverse-mode AD
    Adding cells with centre within boxes 1((-100 -100 -100) (100 100 100))
Solving Linear Equation... 731 s
Main iteration 0 KSP Residual norm 5.879424130218e+00 786 s
Main iteration 100 KSP Residual norm 3.829576270200e+00 894 s
Main iteration 200 KSP Residual norm 3.348720816763e-01 1016 s
Main iteration 300 KSP Residual norm 5.971429321130e-04 1148 s
Main iteration 400 KSP Residual norm 1.522158849654e-07 1302 s
Main iteration 453 KSP Residual norm 5.704343747294e-10 1393 s 
Total iterations 453
Solving Linear Equation... Completed! 1393 s
Residual tolerance satisfied, solution finished!
Computing total derivatives....
Computing total derivatives for fvSource
Calculating dFdField using reverse-mode AD for fvSource
    Adding cells with centre within boxes 1((-100 -100 -100) (100 100 100))
Calculating [dRdField]^T * Psi using reverse-mode AD
Traceback (most recent call last):
  File "/home/uttamcadambi/dafoam_v2/packages/miniconda3/lib/python3.8/site-packages/pyoptsparse/pyIPOPT/pyIPOPT.py", line 210, in eval_grad_f
    gobj, fail = self._masterFunc(x, ["gobj"])
  File "/home/uttamcadambi/dafoam_v2/packages/miniconda3/lib/python3.8/site-packages/pyoptsparse/pyOpt_optimizer.py", line 321, in _masterFunc
    result = self._masterFunc2(*args)
  File "/home/uttamcadambi/dafoam_v2/packages/miniconda3/lib/python3.8/site-packages/pyoptsparse/pyOpt_optimizer.py", line 460, in _masterFunc2
    args = self.sens(xuser, self.cache["funcs"])
  File "/home/uttamcadambi/dafoam_v2/packages/miniconda3/lib/python3.8/site-packages/dafoam/optFuncs.py", line 212, in calcObjFuncSens
    DASolver.solveAdjoint()
  File "/home/uttamcadambi/dafoam_v2/packages/miniconda3/lib/python3.8/site-packages/dafoam/pyDAFoam.py", line 2466, in solveAdjoint
    self.writeFieldSensitivityMap(
  File "/home/uttamcadambi/dafoam_v2/packages/miniconda3/lib/python3.8/site-packages/dafoam/pyDAFoam.py", line 1782, in writeFieldSensitivityMap
    fSens.write("%d\n" % len(sensList) / 3)
TypeError: unsupported operand type(s) for /: 'str' and 'int'
Error evaluating objective gradient at user provided starting point.
  No scaling factor for objective function computed!


+--------------------------------------------------------------------------+
|              Evaluating Objective Function Sensitivities 001             |
+--------------------------------------------------------------------------+
writeDeformedFFDs is deprecated since v3.0.1!
Solution time 1e-08 less than 1e-4, not renamed.

so it is in this function

    fSens.write("%d\n" % len(sensList) / 3)

why is sensList a str? I dont understand. Please advise.

You can see that the adjoint solver converged in this instant but then when the adjoint solution computation restarted, I could not manage convergence.

[raisePiginYourhome]

by the way, I want to know how to visualize the senitivity map through paraview?

[friedenhe]

@uttamcadambi07 Thanks for reporting. The error you saw was caused by a bug in the code. We have just fixed it. Please check this commit: 767274c.

[uttamcadambi07]

Thanks! Updated my code and it works, now I can print the sensitivities using the v2 script.

[friedenhe]

@raisePiginYourhome You can visualize the sens map in Paraview by simply selecting thesens_xxx_xxx variable.

[raisePiginYourhome]

ok, Thank you!