Normalisation by T for Stokes preconditioning

Development
1

Hi all,

Currently findsoln and continuesoln (as far as I know) normalise the search residual by T if they are looking for equilibria.

In the case of Stokes preconditioning [(-L) flag] normalisation should also be avoided because for high T (the intended use case is e.g. T > 10) the pde operator is a T-independent approximation to the steady Navier-Stokes equations. (So if you divide through by T it artificially decreases the search residual.)

I would suggest the following amendment to findsoln.cpp

  -        const bool Tnormalize = Tsearch ? false : true;
  +        const bool Tnormalize = (Tsearch || searchflags.laurette) ? false : true;

plus something similar for continuesoln.cpp.

1 Like
2

Thank you very much for pointing this out.
We will fix this issue.
Mirko