Hello!

Here's the situation : I am simulating the effect of conjugate heat transfer for a carbon rod which has a heated tip (by a laser) and which is parallely located in a gaz flow (hence forced convection).

As a start, I am considering the problem of a flow over a flat plate (but in a 2D axisymmetric system) WITHOUT any energy involved (I am just solving the equations for the flow).

I am interested in the STEADY-STATE of the whole phenomena but I decided to simulate the flow first in unsteady mode until I would obtain a (somewhat) nearly fully-developped flow (it's still oscillating a little but I guess it can't get better in unsteady-mode).

Now I wanted to "finalize" the fully-developped flow in STEADY-STATE to obtain a "real" mathematically fully-developped flow (So that not oscillation at all would occur).

The problem that I get is that as soon as I start my simulation in steady-state (after nearly 10,000 converging iterations in unsteady-state) my residuals starts to diverge (even after 700-800 more iterations).

I don't really like the idea of trial-and-error with the different relaxation schemes, I would prefer to understand why my residuals starts to diverge as soon as I enter the steasy-state mode.

Regards, Dominique P.S.: Any one of you tried some microfabrication simulations with FLUENT (made essentially for the aeronautical field)?

Hi,

when you switch to steady mode, the residuals start to increase, because your flow is fully unsteady. In Fluent they suggest to strt with steady flow and then turn to unsteady mode if the convergence is impossible to achieve.

Start with steady flow, and then eventually switch to unsteady flow.

Bye.

If your flow is unsteady you can't simulate it using the steady solver.

To do steady calculations, you need to be sure of the existence of the steady state.

Hi

ap

Thank you for your suggestions.

1. Yes, there exist a steady state for my model. 2. If I start directly to steady-state, the flow takes an eternity to get fully-developped (if it does at all) and it doesn't give good results. 3. I was wondering that if I starts in UNSTEADY-STATE to nearly fully develop the flow (Which I did and which gives excellent results), then I could use this nearly fully-developped flow in another STEADY simulation...

Nobody has ever done that? Using first UNSTEADY then the STEADY mode (like the actual flow would do...starting from unsteady and eventually stabilized in a steady mode). Using the reverse (STEADY first than UNSTEADY, seems a little illogical to me...)

4. If you have a flow at 0.001m/s that goes from point A (inlet) to point B (outlet) (lets says 1m appart), you will use the STEADY mode to obtain the fully-developed flow? Why not starts with the unsteady mode until the flow is nearly fully-developped and has reach point B THEN enter the steady mode to "fully-developped" it (i.e. to remove the oscillation occuring in unsteady mode)?

Thanks, Dominique

It should be ok to switch from unsteady to steady solutions. If your unsteady solution has achieved steady state, your solution will no longer change as a function of time. Can you tell which residual is diverging (vx, vy, or continuity)? I suggest you start with first order accuracy, which is much easier to converge than the 2nd order. As as suggested, if there is a steady state of your problem, I guess you will get the same solution as the steady state solution. The fully developed boundary is not a problem.

If you switch from unsteady to steady calculation, you just use the unsteady solution as initial condition for the unsteady calculation.

Probably you would obtain the same advantage giving a better initial solution, avoiding the unstedy calculation which usually requires more time.

If your steady calculations give different results if you start from different initial solutions, probably in one of the cases you're not obtaining a really converged solution. Did you reduce the under-relaxation factors? If so, try to increase them a bit and to iterate again.

To simulate a flow that goes from point A to point B, entering the domain at 0.001 m/s:

- use the steady solver if you're interested in the steady solution and you don't need to see the dynamic evolution of the flow. Of course the steady state must exist.

- use the unsteady solver if the system is unsteady by itself or if you want to capture the evolution in the time of the flow.

Hi

ap