[jamescm]

Hi,

I have been using OpenFOAM for several years (v2.3.0 and v7) and have recently stumbled across a strange issue. I am using a custom subsonic reacting flow solver and two cases (5M and 20M tet cells). The issue is related to restarting from the previous time step during reacting flow runs. If I run the case from ignition to, say T10, there are no issues. If I stop the case at an intermediate time, say T5, and then restart the case from T5 and run it to T10, the fields do not look the same. In particular, one field typically has low values (order -5 to -3) and this field is very sensitive to this. I have seen this behaviour in both cases and this did not occur when using OpenFOAM 2.3.0.

I have also run tests including doing a fresh decomposition of T5 in a new directory and running to T10 and also the same test on two different HPCs, which use different compilers and MPI. I am waiting for a local workstation run to finish to see if it an issue with running these cases on a higher number of processors. I have noticed though this issue does not come up in another case I have ran, which used a hex mesh.

If anyone has any advice or has also observed similar behaviour, I would be very grateful for some further information.

Thank you,


J