When I couple the 1D Software GT-Power with StarCD sometimes I get very high oscillations at the interface volumen of GT and Star. I already have testet many ideas, but I still have not found an overall reproduceable error to avoid this oczillations. Some ideas / tricks helps, but not enough.

The application is a intake manifold calculation for normal asperated or turbo charged combustion engines.

It is strange, because this oscillations occur in that time steps when the pressure gradients are very low (means: in that time when the intake valve is cllosed). When the intake valve opens I get a very good high quality devolution / distribution of mass and velocities.

I hope strongle anybody is out there with an advice.

Your Courant number is probably too high at the boundary for the explicit 1D code to be stable. If this is the case, your options are:

1) Decrease your timestep.

This may not be as bad as it seems, because STAR-CD may converge more quickly with a smaller step, only marginally increasing total runtime.

2) Increase the mesh size at the boundaries.

Counter-intuitive perhaps, but it will improve stability. Ideally make the mesh size (in the normal direction) equal to your 1D code's discretization.

Incidentally, what did the GT people recommend?

They recommand exactly what you have sad. Additionally the Flow in the interface volumen has to be straight, means without curvs / bends.

The problem is, that I get oscillations at times of low pressure gradients !!! So decreasing the time step is not a remedy, because the velocity is low.

Do you have similar problems?

I don't actually run GT with STAR-CD, but I do have several years of 1D/3D experience. And in my experience, step-to-step velocity oscillations are always caused by high Courant numbers at the boundary, regardless of local flow conditions. Just for reference, 1D explicit codes are generally less stable at low flows than at high flows (noise takes over).

Out of interest, what is your boundary mesh size (in the normal direction) and what timestep size (in seconds) are you using? Also, how many steps does the 1D code take for each STAR-CD step?

Steve, hold on, I will come back to you.

Steve,

I am still alive. The problem is very important to me, but currently I can't produce results from a test case, because other things have a much more higher priority and time is hurrying, but I will come back to you. Sorry, I hope you will be still here.

So long Hubert

Still watching, but rapidly getting RSI from scrolling the mouse right down here...