[.bastian]

Actually im searching for a way to automatically change the physical timestep in dependency of the actual iteration step.

It should be something like:
its <= 10: 1e-3 s
10 < its 100: 1e-4 s
100 < its: 1e-5 s

Struggeling the forum search i found something which seems to be promising:

Originally Posted by ghorrocks

Have you tried setting the autotimescale factor to a CEL expression then using a CEL expression like if(aitern>15,10,1)?

My solution based on this hint is: if(aitern<=10,1e-3 [s],if(aitern>100,1e-5 [s],1e-4 [s]))

but unfortunately it doesnt work. Setting this as an expression for the physical timestep in CFX Pre the solver starts normally using the set timestep 1e-3 [s] but it doesnt change the timestep when the iteration levels are reached.

Any idea what i did wrong?

thanks in advance

bastian

[ghorrocks]

You can also do this with a 1D interpolation function. I am a bit puzzled about how it did not work, however.

But looking at the bigger picture: I assume this is for a steady state simulation. Decreasing time scale size as you progress to convergence is not recommended. Usually you want to increase the time scale size. Why do you want to decrease it?

[.bastian]

Originally Posted by ghorrocks

[...]
I am a bit puzzled about how it did not work, however.
[...]
Why do you want to decrease it?

Thanks for your Answer.

I also found the variable citern in place of aitern but it doest work either. For the the fact that the expression itself seems to work until it starts with the correct value set for the first ten iterations.
Maybe during the run the expression is not evaluated correctly?!
And yes you`re right, its steady state.

To the more common things.
The simulation itself is a double choking supersonic jet-pump, just to give you a little insight into the model itself.

I'm not sure if its the right explanation until its one of my first bigger projects in cfx but i think you'll know what i mean.
The first few steps, whilst CFX "fills in" the fluid into the domain i want to use bigger timesteps until i see no need to spend more time doing this.

But as soon as the first shock regions occuring it becomes heavily unstable. But all in all it converges using a timestep small enough.

So all in all i want to start with bigger timesteps to get quicker to the point i have to resolve finer.