[streamline90]

Hello guys,

I am working on a steady state simulation involving a flow of air past a high speed train. The object of analysis in particular is the pantograph installed on the top of the train and the lift and drag forces exerted on it.
I have already successfully launched a simpleFoam simulation but the results, in terms of forces, are underestimated in comparison with the experimental data.

You can find attached an image with the case description

Therefore I though I could refine the mesh on the roof of the train to better describe the boundary layer which impacts on the pantograph.
Unfortunately the new mesh would take too much time to be build on the HPC I am actually using, which has a walltime limit of 24 hours.

So I was wondering if I can run two simulations: I start simulating the first half of the train and then I give as inlet boundary conditions for the second half the outlet fluid characteristics found with the first simulation.
Is it possible?

Thanks in advance for your time, and if you have any other strategy to suggest do not hesitate to write me!!

Luca

[nandhakumar]

Hello

I believe you are asking about parallel processing. you can do it with openfoam using decomposeparDict. But you said HPC I have no knowledge in HPC. If you are using normal personal computer you can do it with parallel processing.

Thank you

[streamline90]

Hello nandhakumar,

thank you very much for taking your time to reply me.
Actually I am already running my case in parallel but maybe my initialization of k and epsilon is too small?
In fact in the mail I received notifying that you replied, I read the following answer:

Hello

For your case it seems your initialization of k and epsilon seems very small. Becoz the inlet velocity in at 55.5 m/s believe me it is like a storm. for high velocity flow your initial k and epsilon should be more realistic. Also you said you are using unstructured grid, it needs your schemes should be properly selected. Try using bounded gauss upwind Try modeling with structured grid.

Thank you

I am bit confused dear nandhakumar do you mind taking some time to tell me which is the answer you intended to tell me?

Thanks a lot!!!

[Swagga5aur]

Hello Luca

I believe the utility you are looking for is the mapField utility, a tutorial is located in the tutorial directory incompressible/icoFoam/cavity and a post explaining the utility abit .

As you talk about the issue may be due to capturing the boundary layer may I ask what turbulence model are you using as well as what y+ do you have along the train?

Best regards
Lasse

[streamline90]

Hello Lasse,

first of all thanks for your help.
Answering to your question I am using the k-epsilon SST turbulence model and talking with my colleagues they also suggested me to check the output of the y+
Effectively it seems too high especially in important paches for the boundary layer development such as the train body:

/*---------------------------------------------------------------------------*\
| =========                 |                                                 |
| \\      /  F ield         | OpenFOAM: The Open Source CFD Toolbox           |
|  \\    /   O peration     | Version:  2.3.0                                 |
|   \\  /    A nd           | Web:      www.OpenFOAM.org                      |
|    \\/     M anipulation  |                                                 |
\*---------------------------------------------------------------------------*/
Build  : 2.3.0-f5222ca19ce6
Exec   : yPlusRAS -latestTime -noFunctionObjects
Date   : Oct 10 2017
Time   : 16:57:19
Host   : "node187"
PID    : 17505
Case   : /gpfs/scratch/userexternal/lscorret/KNEE_DOWNSTREAM/WBL_071_ETR1000_SIcalotte_NOcorrugati_BOCCA
nProcs : 1
sigFpe : Enabling floating point exception trapping (FOAM_SIGFPE).
fileModificationChecking : Monitoring run-time modified files using timeStampMaster
allowSystemOperations : Allowing user-supplied system call operations

// * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * //
Create time

Create mesh for time = 3000

Time = 3000
Calculating wall distance

Writing wall distance to field y

Reading field U

Reading/calculating face flux field phi

Selecting incompressible transport model Newtonian
Selecting RAS turbulence model kOmegaSST
kOmegaSSTCoeffs
{
    alphaK1         0.85034;
    alphaK2         1;
    alphaOmega1     0.5;
    alphaOmega2     0.85616;
    gamma1          0.5532;
    gamma2          0.4403;
    beta1           0.075;
    beta2           0.0828;
    betaStar        0.09;
    a1              0.31;
    b1              1;
    c1              10;
    F3              false;
}

Patch 6 named pantograph_collector1 y+ : min: 0.4722751 max: 373.8543 average: 87.84762

Patch 7 named pantograph_collector2 y+ : min: 0.3052835 max: 461.3053 average: 101.2462

Patch 8 named pantograph_horizontality_bar y+ : min: 11.41663 max: 253.7964 average: 78.99773

Patch 9 named pantograph_derivation_arm y+ : min: 0.1030144 max: 289.0489 average: 84.36333

Patch 10 named pantograph_rear_arm y+ : min: 0.3300036 max: 311.8542 average: 52.78361

Patch 11 named pantograph_isolator_derivation y+ : min: 0.04817243 max: 937.1632 average: 60.93609

Patch 12 named pantograph_spring y+ : min: 0.009559944 max: 173.1692 average: 17.38859

Patch 13 named pantograph_lower_frame y+ : min: 0.00012969 max: 263.8775 average: 45.49734

Patch 14 named pantograph_upper_frame y+ : min: 0.07278851 max: 402.0781 average: 69.93129

Patch 15 named pantograph_frame y+ : min: 8.76735e-05 max: 452.2014 average: 56.84387

Patch 16 named pantograph_head_susp y+ : min: 2.933546e-05 max: 479.1439 average: 85.10539

Patch 17 named train_body_housing_3kV_NS y+ : min: 1.409824 max: 1628.665 average: 201.4949

Patch 18 named train_body_housing_3kV_S y+ : min: 0.1940024 max: 878.0415 average: 106.8896

Patch 19 named train_body_housing_25kV y+ : min: 0.4338129 max: 2031.743 average: 174.7839

Patch 20 named train_body_train y+ : min: 0.1223503 max: 7022.329 average: 1487.02

Patch 21 named ballast y+ : min: 49.85643 max: 3265.872 average: 1699.411

Patch 22 named panto_3kV_NS y+ : min: 4.657518 max: 2653.024 average: 296.7346

Writing yPlus to field yPlus

End

At this stage I would like to take advantage of your kindness to ask you some questions:

- This y+ output above shows for every patch the minimum and maximum value of the normalized height from the surface ( i.e. y+ ) of all the cells right on top of the patch surface ( so in contact with the surface )right?

- As I am using wall functions, my y+ should be within the range 30<y+<500. In order to reduce my current value the best strategy would be a refinement on the interested patches, am I right?

Thank you all guys!
Cheers,

Luca

[nandhakumar]

Hello

Sorry that was a reply to some other thread. An error from my side.

Can you upload the case file may be I can take a look.

[streamline90]

Hi nandhakumar,

sorry for the delay. I don't need to bother you anymore cause I found a solution that seems to be more clever to stay into my time limit for the meshing operation.
This is, for the interest of all, running the 3 different phases of snappyHexMesh with three different script.

Thanks again for all your suggestions and have a nice day.
Cheers,

Luca

[Swagga5aur]

Hello Luca, sorry for the late reply.

- This y+ output above shows for every patch the minimum and maximum value of the normalized height from the surface ( i.e. y+ ) of all the cells right on top of the patch surface ( so in contact with the surface )right?

'

Yes its the near wall cell centers which is used, however y+ is not a distance its the dimensionless wall distance consisting of the wall distance y, the fiction velocity and the local kinematic viscosity.

- As I am using wall functions, my y+ should be within the range 30<y+<500. In order to reduce my current value the best strategy would be a refinement on the interested patches, am I right?

Well, the necessary y+ varies with the turbulence model and if you are using a wall function then yes a y+ of 30<500 is right. However, if its the k-omega sst model you are using without a wall function, the y+ should be around to 1.

You mentioned you are using a k-epsilon SST model, and I'm not sure that is right as, to my understanding, the SST model refers to the combination of the k-epsilon model and the k-omega model and is usually called k-omega sst and not k-epsilon sst. I may be wrong just never heard of the k-epsilon sst model.

Regards
Lasse

[streamline90]

Thanks a lot Lasse for clarifying my doubts about y+ !!

Best regards,

Luca