furstj / myfoam Goto Github PK

Dear author:

hello! Your paper has given me great inspiration. But I have a question about the implementation of LU-SGS method in OpenFOAM and another question about the LU-SGS for unstructured grids.

In your code implementation, do the forward and backward step iterations in the LU-SGS method start from boundary elements or internal elements?

Because for structured grids, the Forward and Backward steps start with boundary elements, my question is:
Should forward and backward iterations also start from boundary elements for unstructured grids?

These are my two questions and I look forward to your wonderful answers!
Best wishes!! : )

Hi, Jiri Furst.

I am Jiaqi, from shanghai Jiaotong University, a 5-years phd student. Recently, I have contacted with u in github issues. I am doing my final PhD thesis, which is related to aeroacoustic. Thus, I developed a non-reflecting boundary for high resolution of simulation with LES method.

Your work is quite appealing to me. Its amazing your have done so much of work on dynamic mesh, both DyMFoam and ALE. While it vertified in the 1D sinewave case. I try to explore more possibility, such as 3D case, with rotating blade or oscillating solid. I found you have done some work with it, as shown in your JCP-2018 (the oscillating NACA-0012).

I am emailing to you for asking if you could share this case with me. My idea is to combine it with non-reflecting boundary condition, dealing with the oscillating source of acoustic in the swirl flow duct. Which I have done a lot of theory works and experiments on it. I will introduce your solver in my thesis.

Like u, I am a open source enthusiast, I will share all my phd work in my github in the final. If you are willing, that would be great.

my email: [email protected]

Thanks!
Yours sincerely

Hi, furstj
I am quite interested in your work. I compared rhoPimpleFoam and lusgsFoam solver. The results could be found here.

Some Concerns:

• lusgsFoam with forward ddt scheme is broken, which I cannot comprare with rhoPimpleFoam (forward) case.
• rhoPimpleFoam seems more non-linear than lusgsFoam. see Euler case. （Upper and lower waveform asymmetry in rhoPimpleFoam， but symmetry in lusgsFoam）

Any ideas or suggestions for improving my cases?

Dear Professor Jiri Furst:

My name is Wu Lei, and I am a master student from Xi'an Jiaotong University.Recently, I saw the LU-SGS method you implanted on Github, then I read your paper ”Development of a coupled matrix-free LU-SGS solver for turbulent compressible flows” and I found it great and attractive. I have benefited a lot from learning your codes.

I am very interested in the example NACA0012 in your paper, I am emailing to you for asking if you could share this case with me, I recently want to test different schemes to calculate this case. And I will introduce your solver in my first article.

Thanks!

my email: [email protected]

Yours sincerely

Dear Professor,
I try to use MixingPlane boundary condition in a 2D model as a interface between stator and rotor.
I put MixingPlane condition for patch at rotor side and zeroGradient for stator as indicated in mixingPlaneFvPatchField.H, however the boundary condition has a wrong behavior because it doesn't seem to transfer information upstream as you can see in the picture.

Am I doing it wrong or is there something else?

Best Regards

Hi furstj.
I found that the solver works in the MRF case, could you provide a basic cases? So far I know that for example in "https://www.youtube.com/watch?v=eDUXXcXfX44", public openfoam only offers MRF for incompressible flows (simpleFoam).
My question is how to extend rhopimpleFoam or your solver to Dynamic mesh or MRF?
Thanks~

Hi professor Furstj,
I'm curious about the applicability of myLusgsFoam solver to simulate low-speed flow (almost incompressible).

The troubles are as follows:

Previously, simpleFoam solver was used to solve the low Mach compressor cascade, where the Mach number in the inlet was about 0.067. Using SST turbulence model, the results show perfect convergence.
​Nowadays, I would like to simulate the same case with myLusgsFoam solver with the following boundary conditions:
inlet: fixed value for velocity and temprature, zero gradient for pressure;
outlet: fixed value for pressure, zero gradient for velocity and temprature.

I found that the results simulated by myLusgsFoam solver is more sensitive to the initial field than the rhoSimpleFoam solver.

1. When uniform initialization is used, the pressure field fails to converge for the myLusgsFoam solver, while it converges well for the rhoSimpleFoam solver.
2. Using the results of the simpleFoam solver for initialization, the pressure dimension is modified and the exit back pressure is adjusted. The results also fail to converge for the myLusgsFoam solver, where the pressure field in the channel appears to oscillate back and forth.
3. The flow field is stabilized only with the initialisation of the results computed by the simpleFoam solver. However, the residuals still fail to converge properly, with the 'rhoE' residual fluctuating around 10.

hello,professor Furstj. I've read your paper and I'm very interested in your work.

There's a bit of code in the file forwadSweep.H that I'm confused about, and I've posted below
rRho += (ac + av) * dRhoStar[own] + 0.5 * (rho1*phi1r - rho[own]*phi0r); rRhoU += (ac + av) * dRhoUStar[own] + 0.5 * ( (rhoU1*phi1r - rhoU[own]*phi0r) + (p1 - p[own])*Sf[faceI] ); rRhoE += (ac + av) * dRhoEStar[own] + 0.5 * (rhoE1*phi1r+p1*phi1 - (rhoE[own]*phi0r+p[own]*phi0));

This code should correspond to this formula(show as below)
$$D_i \Delta W_i^{(1)} = -R_i - \frac{1}{2}\sum_{j\in L(i)} [\Delta F_j^{(1)}\cdot S_{ij} + \lambda^*_{ij}\Delta W_j^{(1)}]$$
So my question is, why is this code "+=" instead of " -="? Am I missing something crucial? I look forward to your answer and best wishes : )

Dear professor Furstj,
I am very interested in your work, and I am trying to apply this solver to high-speed compressible flows.
I have tried several cases, including steady-state and transient. However, all these cases crashed, and the results are not satisfactory. I compared the case settings in the tutorial folder and try it to improve my cases, but it doesn't work. Could you give me some advice or suggestions? Thank you for your time and consideration. Attached is one of my cases, shocktube.
Best Regards
shockTube.zip

Hello professor Furstj,
Thanks for professor's last reply. Through the nonuniform List b.c. taht you suggested, the inlet nonuniform total pressure distribution has been successfully implemented in a 3-D high-subsonic compressor cascade. Today, there are several solver setting issues that need your help, due to I am not special in numerical algorithms.
1. I have no idea what the term "internalIterations" refer to in the fvSolution file? Is this the number of Gauss-Seidel sweeps or the maximum number of iterations per physical time step?
2. I want to know whether the "MaInf 0.2" parameter under the "AUSMplusUpfluxCoeffs" dictionary in the fvscheme file has an impact on the calculation results and whether it needs to be adapted according to the inlet Mach number, i.e. "MaInf 0.6"?
3. Is there any experience to share regarding the selection of items "hllcFlux", "betaFlux" and "AUSMplus UpFlux"? I.e., depending on the Mach number range?
4. In my current high-subsonic compressor cascade, the steady calculations based on SST turbulence model show a little unsteadiness. Whether the results can be further converges through modifying the "dbnsFlux" term, i. e. from "AUSMplusUpFlux" to "hllcFlux", and other measures? Or is the transient calculation required?

Looking forward to professor's help!

Hi, I am interested in this LU-SGS solver. But I try to learn your solver and use the case lowMachChannel, there is a error for create thermal object. I use Info to lock the position of bug. It is in createFields.H "autoPtr pThermo(psiThermo::New(mesh));", this order. Cloud you give the reason of this bug? I think it is just the error of create object. The version of openfoam is openfoam-7.

Hello professor furstj,
Recently, I successfully simulated a 2-D high subsonic compressor cascade using your myLusgsFoam solver with uniform total pressure and temperature inlet boundary condition. The "subsonicInletTotal" b.c. is used for inlet velocity boundary.
Nowadays I want to realize a non-uniform inlet total pressure profile along one direction using "codedFixedValue" boundary, where the static pressure is calculated by subtracting the dynamic pressure head (0.5rhomag(U)^2) from the specified total pressure. But I find that there are two forms of density can be access, "rho" and "rho_0".
The question is when the "rho_0" is used, the case can be successfully continued. But when the "rho" is used, the calculation is always divergent. I have no idea for the reason, also I can't find where the "rho_0" is defined. Therefore I hope to get your help. Meanwhile, I want to know whether there is a more reasonable way to define the total inlet pressure profile?
Thanks for professor's help!

Hi, when running the example hypersonicCylinder through parallel, the results diverge.
It's Ok in non-parallel.

 Time = 35
LUSGS: iteration 1
[1] #0  Foam::error::printStack(Foam::Ostream&) at ??:?
[1] #1  Foam::sigFpe::sigHandler(int) at ??:?
[1] #2  ? in /lib/x86_64-linux-gnu/libpthread.so.0
[1] #3  ? in /home/ofuser/OpenFOAM/ofuser-v2006/platforms/linux64GccDPInt32Opt/bin/myLusgsFoam
[1] #4  __libc_start_main in /lib/x86_64-linux-gnu/libc.so.6
[1] #5  ? in /home/ofuser/OpenFOAM/ofuser-v2006/platforms/linux64GccDPInt32Opt/bin/myLusgsFoam
srun: error: cas643: task 1: Floating point exception
srun: Terminating job step 9600744.0
slurmstepd: error: *** STEP 9600744.0 ON cas643 CANCELLED AT 2022-02-14T11:00:02 ***
srun: error: cas643: tasks 0,2-3: Terminated
srun: Force Terminated job step 9600744.0
Execution finished
Script done

Hello prof.,
I would like to ask whether mylusgsfoam can implement the time varying velocity boundary conditions to simulate the effect of upstream wake on a turbine cascade.
Thank you for your help!

Recently, I see that the RANS turbulence model can also be combined with the LU-SGS algorithm in some other codes, which involves jacobi source terms. So I would like to consult with professor to see whether this approach contributes to improving robustness.