Hello there (especially Steve, you are welcome)

Steve: I hope you remember me from last june. Now time has come to take a closer look to the coupling for my intake manifold again, to do a little bit research to get better results. In the meantime my opteron cluster works for the coupling (Gt 6.1 Build 6, and StarCD 3.24)

I still have the high frequency oscillations at low pressure gradients (times when the intake valve is closed already a long time).

I can tell, that the star and gt-model is correctly set up in that way the gt-power tutorial tells me. There are some other settings / improvements from other specialists, which which should help.

At engine speed 6800 RPMs and 1°CA (Crank Angel) timestep (about 2.45E-05 sec) everything is fine, but when I change only the engine speed to 4500 RPMs and again 1°CA (now about 3.7E-05 sec) I got very high oscillations. This oscillation are defnitly not of physical nature.

OK this timestep could be a little bit to large. I have reduced it to 0.5°CA (about 1.85E-05 sec) but there is no effekt to the oscillations. Please remember, that his starcd-timestep is smaler than that of the 6800 RPMs case.

I got it working well with 0.25°CA (about 9.75E-06 sec !!!). But I have other geometries which even don't work with 6100 RPMs and 0.1 °CA !!!.

Now I have the problem, that I don't know in which case 1°CA is suitable or I should better take 0.25°CA.

I suppose there is a (perhaps periodic) dependency between engine speed and starcd timestep.

And I suppose strongly that the problem is NOT (!!!) StarCD, but GT-Power.

I will now start extensive parameter studies to get more information of the oscillation generation.

I think I will write a user-subroutine for star to damp the non physical oscillations by myself.

I hope someone has some good advice.

Hubert Janocha

Without spending too much time on a reply at this stage, I'd suggest three things to add to your study:

1) Investigate how many steps your 1D code is taking for each STAR-CD step (on average). If it's more than 3-4 you will probably get problems at at least some speeds.

2) Check whether your oscillations are from step to step - i.e. +-+-+-+-. If this is also true with a smaller step size, you have classic explicit solver instability.

3) Don't think of your optimum timestep in terms of crank degrees - it neverworks out like that. I would work out an optimum timtestep (in seconds) and the write a user subroutine to shrink STAR-CD's step to some nice rounded value less than that (1/N degrees). E.g. if you find 1e-5 is globally ok and your engine speed is 4000, your subroutine would compute 0.24 degrees as the largest safe step. Then shrink that to 0.2 so that you still hit the degree points. That of course assumes that STAR-CD has access to the 1D model's engine speed (or imposes it?).

Questions:

1) How will you "write a user-subroutine for star to damp the non physical oscillations"?

2) What kind of boundary is causing the problems? Is STAR-CD receiveing mass flow or calculating it?

Hi Steve,

shit happens. I had to interrupt my studies once again, but next week (so god will) I will take a closer look to your advice. I think it sounds very reasonable/auspicious.

I suppose the following: There are two software packages, and for sure every such numerical software package has its own swinging system. Possibly there are interferences, which lead to an amplification of the oscillations at a certain rpm/dt combination. Possibly there there are some other dependencies, which affects the resonance frequence.

1) I have to investigate some effects which can damp the oscillating system, and can manipulate the period. (Perhaps introducing a under relax. fact. for the GT-Star-Interface-volumen, or something like that. Disturbing the period in any way) Currently I don't have a realy good idea for this.

2) I am using a inlet boundary (strongly recommended by GT-Power). In the GT-Manual it is written, that for this StarCD-boundary type GT passes velocity to Star and Star returns the pressure. So StarCD has to calculate the mass flow.

In any case I will tell you my results.

Best Regards, Hubert Janocha

"1) I have to investigate some effects which can damp the oscillating system, and can manipulate the period. (Perhaps introducing a under relax. fact. for the GT-Star-Interface-volumen, or something like that. Disturbing the period in any way) Currently I don't have a realy good idea for this."

In my experience unstable oscillations in coupled 1D/3D models are always of the "in-out-in-out-in-out" variety, with boundary velocity flipping direction from step to step. I'd be surprised if you could damp this with any changes at the STAR-CD side, since their origin is the explicit solution method in the 1D solver.

Anyway, I look forward to an update.