Time step in transient simulation
 

Hi!

I'm having some questions about transient simulations. When performing a transient simulation I get the residuals to decrease to good levels. However they take between 50-100 iterations to get down to these levels and for each new time step they jump up to the same level before decreasing again. I'm guessing that this is because the flow changes quite a bit in the beginning.

Since I'm not interested in the initial stages but instead what happens after the flow has settled I want the simulation to run for a couple of seconds. Initially I use a pretty small time step, around 1e-4 s but because the simulation would take way too long if I kept this time step I increase it after a while.

This is an example of how I vary the time step:

First 10 1e-4 s time steps
Then 10 1e-3 s time steps
Then 10 0.1 s time steps (to "fast-forward" in time)
And then back to 1e-3 s time steps

So the thing I'm wondering is if this kind of "fast-forwarding" is ok when working with transient simulations. When using the 0.1 s time step the residuals are decreasing within each time step but the mean value of the residual is slowly increasing (I don't use a convergence criteria but instead use the maximum number of iterations per time step to control the residuals, I usually set it to 100). I'm using a second order segregated implicit solver with the k-epsilon RNG turbulence model.

Thanks for the help!