Convergence in timestep
 

Hello!

It would be very interesting to hear some different opinions on this issue:

When doing an unsteady calculation, the standard in FLUENT is that convergence within each time step is determined from the values of the scaled residuals, which are scaled by values normally determined at the beginning of the calculation (for example, from acheiveing a steady-state solution to begin from). And these scaling factors do not change when switching between the segregated and coupled solver, for example.

So, I was thinking... 1. Shouldn't scaling be applied within each timestep? Or would that be a too harsh limit on the residuals? 2. Do you generally have to change the scaling factors in any way when switching solver? (I have great difficulties reaching convergence with the coupled solver when starting from a steady-state solution obtained with the segregated solver...)

Thanks in advance!

/Henrik