.. _troubleshooting

Troubleshooting
===============

The number 1 cause of the case crashing is the divergence in the Newton
solver used in algebraic models. Look at the error trace: does the crash occur
in a function called :code:`value` or :code:`derivative`? If so, then it is the
Newton solver.

The divergence occurs due to a bad velocity value being sampled to the wall
model. This typically occurs due to bad (e.g. uniform) initial conditions. The
remedy is either to run your case for a bit without wall modelling to get a
reasonable initial field, or to switch to an explicit wall model, which does not
use a root-finder.

However, often one can also run with :code:`hSampler` set to 0, i.e. sampling
from the near-wall cell. This tends to be very stable.

If your case crashes anyway, try setting :code:`copyToPatchInternalField` to 1
in the boundary condition configuration. This copies the wall value of the
viscosity to the near-wall cell. This is a trick, but it tends to stabilize the
velocity value in the near-wall cell. This should certainly be used if you see
that your wall :code:`nut` values are starting to climb up unreasonably. The
latter is, in fact, how ODE models "diverge", even if the case will be able to
run for a long time.

Finally, don't refine the mesh towards the wall, at least not a lot. This tends
to lead to less stable simulations and won't improve the accuracy.