[melanie]

Hi,

I want to run coodles on an academic case; the goal of the study is to determine the performance of Foam in direct aeroacoustic predition. The case is very simple: the flow is initialized with the velocity to define two corotative vortices in a medium at rest. The solution is known: the vortices are rotating around each other for a number of cycles, then they merge; the expected acoustic frequency is twice the frequency of rotation of the vortices.

I tried to set up the case with coodles; indeed I need compressibility, and as far as I know, LES is the best for this kind of stuff (moreover I want to compare results with Fluent's ones).

I first chose Crank-Nicholson 0.5 but the computation never started (maximum number of iterations exceeded - same problem as in thread OpenFOAM Message Board: OpenFoam: Running / Solving / CFD: Turbulent, compressible and subsonic gas flow - I don't know how to quote the threads).
I then changed to backward (I am chosing the more accurate schemes), and the computation stopped after nearly 600 time steps; meanwhile, there was something wrong as I saw that the vortices were already starting to merge (normally, the merge procees happens after several cycle, each cycle being ~1785 time steps).

I have not yet tried the Euler ddt scheme, but what I would like is some experience with compressible LES cases with Foam. I was told that perhaps it is too restrictive to start with central differencing everywhere, but it is hard to understand what is the best setting for spatial schemes... Which term(s) should I try to tune ?

Thank you very much for freely proposing your powerful code, the more I learn and the more I want to use it!

mélanie

[melanie]

After many tries, I found a "running" setup with Euler time discretization and uncorrected for laplacian and snGrad terms.

The remaining (physical) error I have concerns the boundary conditions. I put pressureTransmissive on the 4 boundaries, but in fact there are still spurious waves entering and disturbing the domain.
I had something similar with Fluent, the sharp corners of the domain were in cause (their implementation of non-reflective conditions is not efficient on sharp corners). Could it be something similar with OpenFoam ?
I have perhaps simply misused the pressureTransmissive BC; here is what I have on each border:
- type patch;
- physicalType pressureTransmissiveOutlet;
- velocity: type inletOutlet;
inletValue uniform (0 0 0);
- pressure: type pressureTransmissive;
pInf 101325;
lInf 0.3;
value uniform 101325;

Could someone give me a hint please on these BC ? if you are interested, I can give you more infos/references or post images of what I am expecting and what I get...

[eugene]

You can only use pressureTransmissive on outlets, but your BC spec looks fine. Try playing around with Linf maybe?

For LES you should be running at least backward differencing.

If you have significant nonOrthogonality in your mesh, you should be running laplacians and snGrad with "Gauss linear limited 0.333" or 0.5.

[melanie]

hi,
concerning the BC, I have done several tries and I am pretty sure that it is somewhat reflective; I'll post some results when I have evidence!
in order to double-check my results, I'd like to post-process the dilatation field which is
theta = div(U) = (dp/dt)/(rho*c0^2)
To do this, divU is already an utility and for the other expression, I need to read rho, what link should I include ? (my case is a compressible one)
I also need to derive the pressure with respect to time, how can this be handled ?
thanks!

[eugene]

You are correct in that it is somewhat reflective. If you know of a boundary formulation that is perfectly transmissive, please let me know.

You can get rho from the thermo model. Check the coodles source code.

[melanie]

Eugene,
For the non-reflective BC, there is first the Thompson formulation, used by Fluent for example, which is based on the characteristic wave relations derived from the Euler equations, cf. [1] and [2]. But I know that this monodimensional formulation is not perfect in the sense that it is very delicate at (sharp) corners, as I have experienced, and problems also occur in the case of oblique waves.

The other formulation, especially designed for aeroacoustics issues, simulates acoustic free field conditions; this method was first described by Bayliss & Turkel [3], and further developped by Tam & Dong [4]. The main difference with the characteristic method is its multidimensional formulation, which can be expressed in polar or spherical coordinates.
These BC have been shown to be more efficient than those based on characteristics methods.

For the implementation though, I don't know the degree of difficulty...

[1] Thompson, K. W., 1987, Time Dependent Boundary Conditions for Hyperbolic Systems I. Journal of Computational Physics, 68, 1-24.
[2] Thompson, K. W., 1990, Time Dependent Boundary Conditions for Hyperbolic Systems II. Journal of Computational Physics, 89, 439-461.
[3] Bayliss, A. & Turkel, E., 1982, Far-field boundary conditions for compressible flows, Journal of Computational Physics, 48, 182-199.
[4] Tam, C. K. W. & Dong, Z., Radiation and outflow boundary conditions for direct computation of acoustic and flow disturbances in a nonuniform mean flow, Journal of Computational Acoustics, 4(2), 175-201.