[srinath]

Hello

I would like to know what the above bc sets?
If i look in the directory for time 0, i find for example in p

outlet
{
type waveTransmissive;
field p;
phi phi;
rho rho;
psi psi;
gamma 1.4;
fieldInf 18855.8;
lInf 3;
value uniform 18855.8;
}
What are lInf, fieldInf, psi.

When is this bc applicable?

[schmidt_d]

See:
http://www.openfoamwiki.net/index.php/HowTo_Using_the_WaveTransmissive_Boundary_ condition

[sxhdhi]

Hi

If I am right, the waveTransmissive BC mentioned above is only for compressible flow.

There is one new bc called convective BC published in forum. But I am not sure we can simply obtained convection velocity needed and seems to setup the value fixed during calculation is not correct.

Anyone has better idea?

Regards

XH

[woody]

Hello David,

I am actually optimizing the waveTransmissive BC by adding a plane wave filter. This filtering works quite fine, but as i am trying to extend the same BC for excitation i really need to understand the equation the waveTransmissive BC is based on...

So far I arrived at some equation that looks somehow like:

dp/dt=(1/dt-Linf*DeltaCoeffs) p(t=n+1) + K/dt pInf

Am I right so far??

Is there any location I can see the NOTES the code refers to??

Regards Tobias

[Bertrand]

Hello,

I would be interested as well... Did u find anything related to this equation in waveTransmissive?

Regards,

Bertrand

[woody]

Hi Bertrand,

I figured it out. Do you have the Paper Boundary conditions for direct simulations of compressible viscous flows
T. J. Poinsot* and S. K. Lelef†
Center for Turbulence Research, Stanford University, Stanford, California 94305, USA
NASA Ames Research Center, Moffett Field, California 94305, USA

Received 23 February 1990;
Revised 12 April 1991.
Available online 24 September 2004.
of Poinsot & Lele

It is simply inserting the two LODI formulations (Eqs. 25+34) by eliminating L1. You receive a differential equation in p. L5 is replaced by K/Linf(p-pinf).
Then you have to reformulate it as a mixed BC.

Hope this helps

Tobias

[Bertrand]

Hello Tobias,

Thank you for the answer.

I have the paper now, and tried to figure out the way you were suggesting.

So, I would arrive at something like

dp/dt + (u1-c) dp/dx1 = K/Linf (p-pinf)

My questions are :

1. In the text, K is a constant : K=sig(1-M^2)c/L, with sig a constant which gives "perfectly non-reflecting" when set to zero, meaning the mixedBoundary should be implemented such as dp/dt + (u1-c) dp/dx1 = 0 ?

2. I bet u1 and x1 are related to the flow speed in one direction. Does that mean that this non-reflective Boundary Condition would be unidirectional or is there a possibility to impose this as the normalGradient mixed Boundary Condition?

3. What is the exact meaning of Linf? The length, counted from the border to specify when the "far field" condition is reached?

I hope I don't bother too much with my many questions ...

Thanks!


Bertrand

[woody]

Hi Bertrand,

I'll try to answer your questions shortly...

1. setting sigma to zero theoretically causes non-reflectivity, but the relaxation to a wanted mean pressure won't take place, this leads to a drift of the mean values. I think Baum et. al computed an optimum sigma...

2. This BC only acts in normal direction

3. Yes, but some sources say you should use the domain length for optimum results. In my opinion you should try to chose a length that correlates with your first instable eigenfrequency.

If you need more info, just ask...

[Bertrand]

Thank you for your insight!

I'll have a look for sigma and try different possibilities for Linf.

How did you implement this Boundary Condition? Because Mixed Boundary Condition is supposed to be steady-state and doesn't allow for time derivative in its expression, if I am not wrong.

I tried to look at it via GroovyBC, but again, the same problem : it looks like there is no other possibility than gradient expression, but no time derivative expression, or...?

The only time derivative expression I could find is actually in convectiveOutlet.

Do you have another Boundary Condition?

Regards,

Bertrand

[woody]

Hi Bertrand,

yes this BC is only for more or less steady state applications. It should also work for smooth time variant systems (a colleague of mine uses it for shock wave simulations).

Yes I do have a better BC, which filters f&g waves, but unfortunately I can't give it to you before we published it. Maybe you should tell me what you want to simulate.

Regards

[Bertrand]

Thank you for the reply,

Yes I understand about the still non published paper.

Well, I don't use OF in any of the "classical" CFD purposes, but to investigate about the frequencies spectrum of a sound produced by the vocal tract (now approximated by a variable cross-section tube), so I use rhoPisoFoam, and definitely need some kind of non-reflective Boundary Condition at the far field, if that makes sense...


Bertrand

[woody]

Hi Bertrand,

this BC should work fine. But I don't see your time variant system (speaking of mean values). Maybe you should try some grid expansion. This should also dampen a lot. The low frequencies then should be dampened by the BC....

Regards

[Bertrand]

Well,

Actually the acoustical results inside the tube are pretty fine (compared to experimental data) but outside the tube is another story. I need to have frequencies from 20 Hz until approx 6000 Hz, and therefore, the numerical schemes are too dispersive and dissipative outside my resonating cavity : damping subtle acoustic waves I'd like to study. Hence, I would need a non-reflective BC, of the type given in the paper from Stanford, mixing a normal gradient and a time derivative at the boundary of my domain.

Regards

[woody]

But still, your simulation is not time variant in the sense of means...

Or am I missing something?

[Bertrand]

Well,

my mistake : I forgot to mention I am using a pulsating parabolic inlet, so time-depending (cf. infra) ...

inlet
{
type groovyBC;
variables "A=25;f=50;xp=pos().x;yp=pos().y;X=max(xp);Y=max(y p);";
valueExpression "( (sin(2*pi*f*time())>0) && (sqr(xp/X)+sqr(2*yp/(Y*sin(2*pi*f*time()))))<=1) ? vector(0, 0, ( A*(1-(sqr(xp/X)+sqr(2*yp/(Y*sin(2*pi*f*time()))))) )) : vector(0, 0, 0)";
}

[woody]

Still, not in terms of means....

[Bertrand]

Well I could give it a try but I am not interested in mean values.

[ptbs]

Originally Posted by woody

Hi Bertrand,
3. Yes, but some sources say you should use the domain length for optimum results. In my opinion you should try to chose a length that correlates with your first instable eigenfrequency.

Hi Woody

I am looking for Wave Transmissive BC parameters to suppress instabilities under rhoPorousSimpleFoam. I have 2 questions:

1) considering your remark and according to your experience, do I have to understand that we have to try reduce Linf until oscillation occurs and than just increase it a bit to eliminate reflective waves or is there more subtle rule to apply to minimize the impact of Far Field position on modifying the actual value of P at the outlet?

2) Is it normal (as generally described in tutorials or reported examples) to apply in the dictionary the same pressure for outlet and Far Field position? What is the meaning?

Best regards

[woody]

Hi Patty,

my experience might seem to be greater than it really is, but I'll try to report it to you. For the sake of sense I start answering your second question first:

2) Is it normal (as generally described in tutorials or reported examples) to apply in the dictionary the same pressure for outlet and Far Field position? What is the meaning?

As the far field value is rather a stabilization point for the oscillatory parts than a far field value the mean flow should adapt to this value. Therefore the far field value should not differ from the steady state value for standard problems.

1) considering your remark and according to your experience, do I have to understand that we have to try reduce Linf until oscillation occurs and than just increase it a bit to eliminate reflective waves or is there more subtle rule to apply to minimize the impact of Far Field position on modifying the actual value of P at the outlet?

The first idea is quite sensible, the second one drops out due to 2). If you are not so interested in keeping the mean value, you might further reduce the Linf in terms of orders of magnitudes. Additionally you should think about buffer zones (increasing cells), artificial viscosity to reduce your reflection.

Hope this helps

[ptbs]

Originally Posted by woody

Additionally you should think about buffer zones (increasing cells), artificial viscosity to reduce your reflection.

Hi Woody
Thanks for your explanation. Just one more precision: to have "buffer zones", I suppose you mean increasing the NUMBER of cells close to the outlet? It it so?

Best regards