[hmasenger]

hi dear friends
i have a dumb question but i cant fined the final answer for it in the CFX manual.
the question is"when to decrease the time step and when to increase it"
how can i estimate a proper initial time step?
and one more question which is "what is the difference between the inner loop and outer loop in an steady state model?"
any reference would really help

regards

[brunoc]

Hi Hamed,

The CFX documentation has an extensive text on that. Take a look at section 15.4.1:
CFX > Modeling Guide > Advice on Flow Modeling > Timestep Selection > Steady State Time Scale Control

Cheers

[hmasenger]

thanks brunoc I will take a look at that but what about the inner and outer loops in a steady state problem? can you explain it a little in a simple way ?

[brunoc]

Maybe I'm not understanding your question correctly, but inner and outer loops are only present in transient simulations, not steady state ones (unless we're talking about 2-way FSI, but that's another subject).

- The inner loops (or coefficient loops, as CFX calls them) are iterations within a timestep in order to correct the non-linearities for the result of that given time.

- The outer loop are the iterations that advance the simulation in time.

So, if you have a timestep of 0.1 s, the outer loops are the ones that advance your simulation from 0.1 to 0.2, 0.3, etc. The inner loops are the ones that solve your current timestep until it reaches a desired residual level.

[hmasenger]

thanx bruno. you have answered my question .but there is one more question.
if my model needs to be solved as transient to reach a final steady state condition then does it really matter how many inner loopes hav been set? for example if the steady state model converges after 100 iterations and i want to run a similar model in transient(which i know it is going to be steady sate after some time steps), then i should set the inner loops to 100 or it is enough to set it to 5-10 by CFX default?

[singer1812]

Why does your model need to be solved as a transient to reach the SS value?

Under most conditions, it only needs to be done that way if you are interested in some solutions prior to reaching SS.

[hmasenger]

Originally Posted by singer1812

Why does your model need to be solved as a transient to reach the SS value?

Under most conditions, it only needs to be done that way if you are interested in some solutions prior to reaching SS.

i am modeling 2 phase flow to study cavitation phenomena.i figured it out that in some cases a SS model is not converging and a transient model should be used

[singer1812]

If you find you need to run it transient to get it to converge, then it probably has highly transient effects in it and most likely isnt really a steady state problem, which in turn means to get an accurate solution, you will have to resolve the time.

If it isnt highly transient, then it probably means you just havent set your timescale properly in the steady state solution to get it to solve.

[hmasenger]

Originally Posted by singer1812

If you find you need to run it transient to get it to converge, then it probably has highly transient effects in it and most likely isnt really a steady state problem, which in turn means to get an accurate solution, you will have to resolve the time.

If it isnt highly transient, then it probably means you just havent set your timescale properly in the steady state solution to get it to solve.

i used automatic time scale in SS model and as i know(pleas confirm if i am right) changing the time scale just effects the convergence speed(rate) but doesn't cause convergence !

[singer1812]

Autotime scale is CFX estimate on what to use.

There is a scale factor on autotimescale that you can set, which will adjust CFXs calculated timestep by that factor.

Also you can use local, or physical time step, for greater control.

Using these different types of scales/steps can get you through some tricky physics that might occur as the problem is setting up towards steady state.

Just setting autotime step doesnt mean CFX is going to move happily to convergence.