CFD Online Discussion Forums

CFD Online Discussion Forums (https://www.cfd-online.com/Forums/)
-   CFX (https://www.cfd-online.com/Forums/cfx/)
-   -   is it possible to predict how long it takes to reach steady state solution in unstead (https://www.cfd-online.com/Forums/cfx/126206-possible-predict-how-long-takes-reach-steady-state-solution-unstead.html)

Alimohamadi_nasr November 11, 2013 03:26

is it possible to predict how long it takes to reach steady state solution in unstead
 
is it possible to predict how long it takes to reach steady state solution in unsteady approach based on the residence time? (my favorite monitors shows constant or periodic shape)

For Example if residence time of a problem is 7*10^-3 [s], and it will be solved unsteady, after how long it will become steady if it has steady state solution.

I know it depends on the initial condition, but I want an estimation for it.
for example if its initial condition is its steady state or if we start from the first point?

I have a problem that is in laminar regime and I think its impossible to solve it in steady state mode, and papers also have unsteady approach, to have convergence. I choose dt=2.5*10^-7 [s] and it becomes converged in each time step (10^-5 for continuity and 10^-7 for velocity), it is 3D and I have started from steady state solution as initial condition. I want to know after how many time steps it becomes steady. I know it has steady state solution.

it is kind of instability flow.

Thank you very much

ghorrocks November 11, 2013 05:18

No, you cannot predict in general how long steady state will take at all, let alone based on residence time. Different classes of flows require different times so obviously different times are required.

You might be able to estimate how long it will take for your class of flow, but it will only work for your class of flow. It will not be general.

Note that if you are marching a transient simulation out to steady state then you do not care too much about the time history of the flow. So use first order time differencing and do not converge the timesteps too tightly. Loose convergence is OK, and then go tighter as you approach your final solution. You just want to quickly and approximately march it out in time and only worry about accuracy at the end.

Alimohamadi_nasr November 11, 2013 05:28

Thank you very much for your consideration.

Alimohamadi_nasr November 11, 2013 06:46

only one another question that may be easy for you, my geometry has one converging Channel that flow comes through it and it divided to two diverging channel with ratio 1:9 (10 percent goes from one of diverging channel and 90 percent goes from another) and its laminar flow. because of pressure gradient there is separation i and vorticity is produced in diverging channels.

my question is that, even in this case, is it OK to change form high order discretization to first order and after approaching to convergence again change it to higher order discretization.

ghorrocks November 11, 2013 07:11

In the vast majority of cases it does not matter how you approach convergence. As long as the final run to full convergence is done with accurate settings then the final result will be good.

In very rare cases you need an accurate approach to convergence as well. An example would be a diverging channel where the jet can stick to one wall or the other (due to the Coanda effect). If your approach to convergence was nto accurate the jet could stick to the wrong wall and the final run to convergence could be on the wrong wall and it woudl stay there. But this sort of bifurcation is rare.


All times are GMT -4. The time now is 07:09.