[sathya123]

Hello foamers,

I have lately been testing the compressible LES module in foam-extend-3.2 coupled with the dbnsTurbFoam solver, on simulating the flow field through the ONERA S8 transonic channel.
I have been facing some problems with the numerics, which I would like some clarity on. My final aim is to use the validated module to conduct some turbo-machinery simulations.

The test case:
The test case comprises of a straight channel with a bump on the lower wall, which accelerates the flow to supersonic speeds at the throat of the channel. An imposed static pressure at the outlet coupled with total pressure boundary condition at the inlet regulates the mass flow at the inlet of the channel.
A normal shock is formed at the throat of the channel, introducing and exacerbating boundary layer thickening, eventually leading to flow separation. A 2 way shock boundary layer interaction is therefore triggered, resulting in a characteristic lambda shock pattern.

URANS results
I ran 2D simulations with the k-omega SST model , with a turbulence intensity of 10% at the inlet. I then obtained the mean flow field averaging over 3000 time snapshots. I have enclosed a picture of the mean fields. I resolved the boundary layer, to a yplus ~20, and maintained near wall aspect ratios to within 10 in the vicinity of the throat. I had discretized the Taylor micro-scale along both the x and y directions. All the advective terms of the N-S equation were discretized with the Rusanov flux coupled with the standard Barth-Jespersen limiter. All viscous fluxes were discretised with the Gauss linear scheme with non-orthogonal corrections.

Initialize the 3Dles module
I initialized the 3Dles fields with the fields of the last time step of the U-RANS module. I discretized the Taylor microscale in the thrird direction as well and maintained near wall maximum aspect ratios around 10, in the lambda shock region.
I used the one equation turbulent kinetic energy transport model to solve the sub-grid scales. I have enclosed the fvSchemes and fvSolution files.I used an inlet boundary condition for the sub grid scale "k" =2 e-05.
Besides being very time consuming (limited to using fixed time step of 1e-08, from stability constraints), I also faced numerical issues with the one equation transport equation solver, which I paste below:

Time = 0.00091075
DILUPBiCG: Solving for k, Initial residual = 1.95254e-05, Final residual = 3.86969e-09, No Iterations 1
ExecutionTime = 83797.4 s

Courant Number mean: 0.0022682 max: 0.0541504 velocity magnitude: 498.586

Time = 0.00091076
DILUPBiCG: Solving for k, Initial residual = 1.95273e-05, Final residual = 3.86684e-09, No Iterations 1
bounding k, min: -9.9277e-05 max: 176780 average: 50.4986
ExecutionTime = 83800.2 s

Courant Number mean: 0.0022682 max: 0.0541395 velocity magnitude: 498.634

Time = 0.00091077
DILUPBiCG: Solving for k, Initial residual = 1.95287e-05, Final residual = 3.86141e-09, No Iterations 1
bounding k, min: -8.01284e-05 max: 176948 average: 50.4987
ExecutionTime = 83803.1 s

Courant Number mean: 0.0022682 max: 0.0541287 velocity magnitude: 498.677

Time = 0.00091078

As one can see, unrealistic values for the turbulent kinetic energy are obtained, (negative k and very large values of the sub-grid scale turbulent kinetic energy).
I have set an under-relaxation factor of 0.3 for the k-transport equation.

Could someone please point out where I am going wrong with the simulation setup? Any advise or inputs?

Thank you!

[sathya123]

Hi all,

Has anyone validated the compressible LES module with dbnsTurbFoam? Any inputs?

Thanks again!

[sathya123]

Could this be a problem of time step? But as one can see, I use a very small time step of 1 e-08, and the maximum courant number does not cross 0.1

[sathya123]

I managed to extract the results at the time when the turbulent kinetic energy began to start getting unbounded.

I have enclosed snapshots of the velocity magnitude at the instant. As one can see, the point of separation is predicted quite upstream as compared to the RANS simulations.

Moreover, I have enclosed a plot of the SGS turbulent kinetic energy at the snapshot. Is it realistic to have such large values of the sub-grid scale turbulent kinetic energy.

Thank you,

[sathya123]

Has dbnsFoam been tested with the compressible LES solver yet? Anyone? I would be grateful to receive any advise/comments,

Thanks!

[sathya123]

To summarize my LES simulation,

1. Low value of the sub-frid scale 'k' applied at the inlet
2. Aspect ratios of 10:10:1 in the block containing the lambda shock, maximum aspect raio of 100:10:1 in the sponge block
3. Compressible LES wall functions used, since yplus resolved to 20.

[sathya123]

I was just checking a post on LES, I found this:

The choice of the scheme for LES should be based on the numerical diffusion it introduces. Ideally you should use linear, or cubic. You need good and prett uniform mesh anyways for LES, since you are assuming the filter operator and the differential operator are commutative, which is not true on non-uniform grids!
The filtered schemes are a last resort scheme in my view, if you can't really do better with your mesh. Upwind and linearUpwind, but also limitedLinear and QUICK should be avoided because they are too dissipative.

For the time scheme, backward is the prmary choice among the schemes available in OF. The constraint on he time step is generally not an issue in LES because you are tied by the requirement of resolving the flow large-scale time scales.

I was wondering if numerical errors could be introduced from non-uniformities in the mesh? My case has quite a non-uniform mesh, especially with near wall aspect ratios of 10.
How would I use the filtered schemes? How do they work? Has anyone tested them on compressible flows?