CFD Online Discussion Forums

CFD Online Discussion Forums (http://www.cfd-online.com/Forums/)
-   Main CFD Forum (http://www.cfd-online.com/Forums/main/)
-   -   Initial value to accelerate a simulation (http://www.cfd-online.com/Forums/main/113345-initial-value-accelerate-simulation.html)

The_Architect February 18, 2013 10:15

Initial value to accelerate a simulation
 
Hello,

I'm working with a project where we're trying to figure out how much an "appropriate" initial value would save computational effort during a simulation. We will run simulations under different conditions. However, I find it hard to find any literature about the topic. Does anybody have an idea where I could read about the effect the initial value has on the convergence of a simulation?

Thank you!

andy_ February 18, 2013 11:32

Quote:

Originally Posted by The_Architect (Post 408461)
I'm working with a project where we're trying to figure out how much an "appropriate" initial value would save computational effort during a simulation. We will run simulations under different conditions. However, I find it hard to find any literature about the topic. Does anybody have an idea where I could read about the effect the initial value has on the convergence of a simulation?

Are these steady-state or time accurate simulations?

The broad answer is not much so long as you avoid a big bang. I don't think it is much of an academic topic but more practical knowledge one tends to pick up after doing a few CFD simulations. If your flow is prone to big bangs (e.g. aggressively diffusing flows) then turning on the inlet steadily over a few steps rather than instantly can help quite a lot, as can starting from a previous solution or, for transient flows, imposing a tight tolerance for mass conservation on the first iteration to get the flow moving everywhere although this can sometimes backfire if the generated flow is grossly unrealistic and takes a long time to blow out the exit.

The_Architect February 22, 2013 06:56

Thanks for your reply Andy.

To me more precise, I'm looking into ship hull simulations. "Big bangs" do happen at the early stages of simulations, the free surface and the wake both create problems. My idea would be to use a previous flow field of a hull as the initial value (MapFields in OpenFoam) for the simulation of a new ship hull. However, as this is an academic work, I am still looking for any kind of references about the topic.


All times are GMT -4. The time now is 18:09.