[artkingjw]

It seems like the solution is to change the back flow specification from 'normal to boundary' to 'neighboring cells'. This way the back flow along the top face eventually dies back down to 0.1% or 0.0%, and the flow field looks normal again, converging to E-6.

It's still weird why it works for Mesh 3.9 but not 3.10...

[artkingjw]

Originally Posted by artkingjw

It seems like the solution is to change the back flow specification from 'normal to boundary' to 'neighboring cells'. This way the back flow along the top face eventually dies back down to 0.1% or 0.0%, and the flow field looks normal again, converging to E-6.

It's still weird why it works for Mesh 3.9 but not 3.10...

The alternative is to make the domain absolutely huge. I tried a new mesh that's twice the height of Mesh 3.10 (which is already ~ 40-50 deltas), and that mesh does not report reverse flow issues.

[LoGaL]

I am a bit puzzled by it as well, because when I looked at the size of your domain, it seemed large enough