Natural convection temperature divergence

[artus_ns]

Hello,

I am trying to simulate natural convection around some heated components in a closed 1m3 volume. The problem I have is temperature divergence. The initial and ambient temperature of the surrounding air is 70° and the components are at 125° and 170°. But at the end of my calculation there are a small number of elements located in the rising air that are below 70°, which is not physical.

I see two potential sources of this problem:
1) I have used too larger time step (0.5s)
2) The Linear production k-epsilon turbulence model is not appropriate for my mesh, hence warning 'mesh too refined at the wall'

Any advice would be greatly appreciated.

Many thanks,
Nico

[Antech]

Hello.

For natural convection cases SST turbulence model is usually used, k-epsilon is definitely not a good choice for this. Your mesh should have a boundary prism layer at the heated surface with Y+ around 1.0 (see, for example, Y+ calculator at cfd-online.com, let the freestream velocity be 1.0 m/s, then verify in ParaView).