Hi all, maybe we found anothe
 

Hi all,
maybe we found another bug in the 1.4.1 version.
Using a LowReynolds turbulence model such as Launder&Sharma it is impossible to set k=0 at wall even if this is the proper boundary conditions.
The error is related with the two source terms with k_ at the denominator. We never had such problem in version 1.3: it works perfectly.

Find attached an easy case to be run with simpleFoam, the turbulence model is the Launder&Sharma with fixedValue for k equation at the wall. If k = 0 version 1.4.1 blows up while 1.3 works perfectly. If k is set to one both versions runs.

cylcicSimple

Please let us know if we are right or there is something missing in our set up.
Thanks
Luca & Cosimo

Division by zero is certainly
 

Division by zero is certainly a problem and I would expect it to be so in 1.3 and 1.4.1. The real issue is that in all versions of OpenFOAM so far the sources are volFields including boundary values which are not used and in this case the source boundary evaluation is causing a problem. For the moment the only solution is to set k on the boundary to be a small but non-zero value.

In the next version of OpenFOAM the sources will be evaluated as DimensionedFields which do not include the boundary values which is both more efficient and in this case more consistent.

Hi Henry, we agree with you:
 

Hi Henry,
we agree with you: division by zero is a problem in 1.3 and 1.4.1 as well. But when you say: "boundary values which are not used" it seems to us that they are really not used in version 1.3 but in 1.4.1 they are, we again agree with you boundary values whould not be used.

We have done quite a lot of simulations with LowReynolds meshes having no problem in putting k = 0 at the wall. So we do not really understand what changed in the source term treatment and why.
Is it something related to the new class FvsPatchFields (we are not really into them yet)?
Can you please check if the case is running on your 1.3?

Thanks again for your explanation

Luca & Cosimo