Debugging a Bug in Parallel Boundary Values in DASpalartAllmarasFv3FieldInversion for Feature Extraction

[lnhang0618]

Dear Prof He,
I encountered a bug while modifying DASpalartAllmarasFv3FieldInversion to output a feature named PbyD, defined as: PbyD_ = Cb1_ * phase_ * rho_ * Stilda * nuTilda_ * betaFieldInversion_/(Cw1_ * phase_ * rho_ * fw(Stilda) * sqr(nuTilda_) / sqr(y_));. This feature is crucial for my machine learning analysis. However, when running the script with mpirun -np 16 python runScripts_v2.py, I observed that the processor boundary values, such as proboundary0to1, return value uniform inf. This issue prevents reconstructPar from functioning properly, which is unusual. I confirmed that other features are correctly output and that this issue does not occur in non-parallel execution.

My questions are:

1. Could the betaFieldInversion_ term be causing errors at the parallel boundaries?
2. Are there alternative methods to obtain feature values in DA models in parallel execution?

[lnhang0618]

DASpalartAllmarasFv3FieldInversion.zip

[friedenhe]

1. It should not be related to the betaFieldInversion_ term.
2. Can you try to do `volScalarField PbyD("PbyD", nuTilda). This will first copy the internal and BC values from nuTilda to PbyD. And then, you can assign the values to PbyD.

Also, please call PbyD_.correctBoundaryConditions() after you assign new values to it.

[lnhang0618]

I have fixed the issue based on the suggestion!