CFD Online Discussion Forums

CFD Online Discussion Forums (https://www.cfd-online.com/Forums/)
-   Main CFD Forum (https://www.cfd-online.com/Forums/main/)
-   -   transient behavior dependent on time step (https://www.cfd-online.com/Forums/main/15231-transient-behavior-dependent-time-step.html)

Fab May 27, 2008 15:53

transient behavior dependent on time step
 
Hi,

I wonder, if one can expect that the transient behavior at a certain probe location is dependend on the chosen time step!? E.g. would a time step size of 0.01sec compared to 0.001sec change the output, when watching the low frequencies around 1Hz? I assume that the output depends more on the numerical diffusion, than on the time step size as long as it runs stable!?

Thanks. Fab

Vinayender May 29, 2008 09:04

Re: transient behavior dependent on time step
 
Hi Fab,

When you say that total error depends only on numerical diffusion rather than on time step... Rember when you see MPDE (Modified Partial Differential Equation) for your PDE you see there are spatial derivatives and temporal derivatives in the Trunction Error. That is the error depends not only on the spatial descritisation but also on the time discritisation....As the time step tends to very small your error corresponding to temporal discritisation goes to zero then your error will corresponding to only sptial discritisation....This is all valid for only Unsteady flows...

For steady flow, your final steady state solution will NOT depend on your time step provided your scheme is stable and the things will be right as said by you (i.e, the error is because of only spcatial discritisation )

Thanks, Vinayender.

Fab May 29, 2008 12:38

Re: transient behavior dependent on time step
 
Hi Vinayender,

thanks, o.k. I understand this. So actually the time step size determines the diffusion as well and has the same influence as turbulent or viscous diffusion; and the diffusion is proportional to the step sizs. Then it would not be enought just to check the CFL number due to some osziallations, but needs to calculate the diffusion as well!?

Thanks! Fab


All times are GMT -4. The time now is 20:54.