[spl]

Hi

I'm running a transient simulation for a radial turbine and I would like your advice on time step size and number of iterations per time step. A number of studies have used times steps range from 0.6 degrees per time step to 6.5 degree per time step but very few have stated how many iterations are used per time step. My understanding is that less iterations are required when using smaller time steps. So I have a few questions -

Is there a general rule for number of iterations per time for CFX?

What is the best method of judging convergence within a time step?

Is it best to select a number of iterations per time step (say 20) and then reduce the time step until the transient results are consistent or is there a better approach?

Thanks

[ghorrocks]

Is there a general rule for number of iterations per time for CFX?

Yes. 3-5 for most simulations, 5-10 for simulations with tricky coupling (eg some mutiphase simulations)

What is the best method of judging convergence within a time step?

The residuals, and in some cases imbalances. In very rare cases other criteria are needed as well.

Is it best to select a number of iterations per time step (say 20) and then reduce the time step until the transient results are consistent or is there a better approach?

There is a better approach. Use adaptive time stepping to home in on 3-5 coeff loops per time step. Then it will automatically find the right time step, and change it as the simulation progresses.

[JuPa]

Originally Posted by ghorrocks

The residuals, and in some cases imbalances. In very rare cases other criteria are needed as well.

It's extremely difficult to converge every time step within 1% imbalance.

[ghorrocks]

That depends on the simulation. Some are easy some are hard. But the bigger question is whether it is required for the simulation at hand. If it is needed then you have to do it regardless of it being hard.

[spl]

Thanks for your help.