[MVRC] Mantium Virtual Racecar Challenge 2019 (Grand Prix Cars)

[CAEdevice]

Thanks Rjsa. This is what I do when I decide to longer iterate the simulation while the simulation is running. But what I have to do after the simulation has been completed to resume it? Maybe I should re launch it? Let us see tomorrow, I don't have a workstation available today.

[yinlad]

Hey all, I'm so sad I only just found out about this! Will definitely be fully involved next season, and hopefully give the last event of 2019 a bash too. Some really interesting stuff going on here

Cool! but you’ve only got one week to design your car for the last event of 2019!

Thanks all, yeah only a week I know. We're not too worried about being competitive as much as learning the tools at the moment. Would be nice to be able to enter at least the one event though, even if we're stone last. I'm currently working with the car that you have kindly posted machin, just to get used to things!

[machin]

It might be a good idea to download CAEDevice’s wings too!

[rjsa]

Thanks Rjsa. This is what I do when I decide to longer iterate the simulation while the simulation is running. But what I have to do after the simulation has been completed to resume it? Maybe I should re launch it? Let us see tomorrow, I don't have a workstation available today.

That’s the information I miss too, it will always start from 0.

[wb92]

I got this from Andre a while back but didn't manage to get it right:

We can get a running standard MVRC simulation and make run further even if it was finished?

After it is finished you have to edit the controlDict file. Change the number of endTime (this is the number of total iterations), the writeInterval and change startFrom startTime; to startFrom latestTime;

Thanks Rjsa. This is what I do when I decide to longer iterate the simulation while the simulation is running. But what I have to do after the simulation has been completed to resume it? Maybe I should re launch it? Let us see tomorrow, I don't have a workstation available today.

That’s the information I miss too, it will always start from 0.

With bit of experience in OpenFOAM - maybe that will help you, but I would take confirmation from LVDH/Andre

I've noticed that MantiumFlow has some unusual for me ControlDicts and it switches simulation from first order to second relatively to max number of iterations, so I would check first what will happen exactly when your simulation would be restarted.

However with regular OpenFOAM cases you can continue already finished cases with changes in controlDict, but I have completely no clue how post-process will work with 2nd set of data, as new run saves new files with new calculated data, eg. forces on surfaces.

Let's say you want to continue calculation after 2000 iterations and run additional 500 iterations (2500 in total):
your controlDict will look like that

• application simpleFoam;
  startFrom startTime;
  startTime 0;
  stopAt endTime;
  endTime 2000;
  deltaT 1;
  writeControl timeStep;
  writeInterval 2000;
  purgeWrite 0;

You need to change it to:

• application simpleFoam;
  startFrom startTime;
  startTime 2000;
  stopAt endTime;
  endTime 2500;
  deltaT 1;
  writeControl timeStep;
  writeInterval 500;
  purgeWrite 0;

I left purgeWrite visible, as this one is good one when you want to do tests of custom settings for the simulation. This parameter tells you how many last writeIntervals will be saved, when "0" all intervals will be saved, when "1" only last, when "2" last, and 2nd last and so on.
On previous example if you would run from 0 to 2500 iterations with intervals of 500 iterations you will have 5 timesteps saved beside "0": 500, 1000, 1500, 2000, 2500.

It may happen that simulation could converge before your endTime, then you need to lower the residuals control in system/fvSolution file from e-5, to e-6 or lower:

• SIMPLE
  {
  nNonOrthogonalCorrectors 1;
  residualControl
  {
  p 5e-5;
  "(U|k|epsilon|omega)" 1e-6;
  }
  }

Such convergence happens mostly on first order calculations, second order to properly converge requires much more detailed mesh, iterations, better model, etc. It's always a trade-off between numerical convergence, stability, speed of calculations and if for real case numbers match reality or at least match trends.

That should work, if anyway somewhere I missed something regarding typical OpenFOAM case, let me know. Still some details may differ due to MantiumFlow customizations.

BTW
I couldn't manage 'code' flag to make proper rows with code parts that why I used 'lists'

[maunde]

Hello everyone,

Thank you very much for posting all this informative information. I am interested in participating in the MVRC next year, and would like to know more about how the competition works. How are the HXs modelled? Is it with a porous medium? Are we given the details of the HX for simulation purposes?

How does the simulation of the cars work in deeming the standings for each race?

Thank you in advance for any advice,
Maunde

[LVDH]

• application simpleFoam;
  startFrom startTime;
  startTime 0;
  stopAt endTime;
  endTime 2000;
  deltaT 1;
  writeControl timeStep;
  writeInterval 2000;
  purgeWrite 0;

You need to change it to:

• application simpleFoam;
  startFrom startTime;
  startTime 2000;
  stopAt endTime;
  endTime 2500;
  deltaT 1;
  writeControl timeStep;
  writeInterval 500;
  purgeWrite 0;

Almost right:
you need to change from this:

• application simpleFoam;
  startFrom startTime;

to this:

• application simpleFoam;
  startFrom latestTime;

Hello everyone,

Thank you very much for posting all this informative information. I am interested in participating in the MVRC next year, and would like to know more about how the competition works. How are the HXs modelled? Is it with a porous medium? Are we given the details of the HX for simulation purposes?

How does the simulation of the cars work in deeming the standings for each race?

Thank you in advance for any advice,
Maunde

You should simply sign up, download the demo and try to run a first simulation. You will see how the case is set up, what the porous parameters are (they could change next year) and after running a simulation, you will see that MFlow will export aerodynamic values that can get copy-pasted into the virtual stopwatch, which calculates the performance parameters into a lap-time which is used to determine the race winner.

[CAEdevice]

Thanks wb92, Rjsa, LDVH.

This is my strategy: I launch a simulation with the standard MVRC template (it lasts about a night, I am using an old 2013 workstation this year) and, only if the results are promising, I would like to add 500-1000 iterations to be sure about the convergence (in the last race my cop changed significantly from 2500 to 4000 iterations).

Next year I hope to be back to the brute force of the workstation I have in office.

[wb92]

Next year I hope to be back to the brute force of the workstation I have in office.

what kind of specs that workstation has?

[yinlad]

You should simply sign up, download the demo and try to run a first simulation. You will see how the case is set up, what the porous parameters are (they could change next year) and after running a simulation, you will see that MFlow will export aerodynamic values that can get copy-pasted into the virtual stopwatch, which calculates the performance parameters into a lap-time which is used to determine the race winner.

Hey LVDH what would you say a ballpark time frame should be for a MVRC_Fast sim to complete in MantiumFLOW should be? I get that depends on the PC but for a fairly high spec workstation, are we talking minutes, hours, days etc.

Cheers

[CAEdevice]

You should simply sign up, download the demo and try to run a first simulation. You will see how the case is set up, what the porous parameters are (they could change next year) and after running a simulation, you will see that MFlow will export aerodynamic values that can get copy-pasted into the virtual stopwatch, which calculates the performance parameters into a lap-time which is used to determine the race winner.

Hey LVDH what would you say a ballpark time frame should be for a MVRC_Fast sim to complete in MantiumFLOW should be? I get that depends on the PC but for a fairly high spec workstation, are we talking minutes, hours, days etc.

Cheers

I think MVRC_fast can be completed between 2 and 4 hours.

[CAEdevice]

Next year I hope to be back to the brute force of the workstation I have in office.

what kind of specs that workstation has?

An old (2013) Intel i7, with a huge quantity of RAM but a very slow clock (slower then the original clock, I now use it as a backup server and it has temperature limits).

I also have a more modern and powerful workstation in office, but recently I had to dedicate it to a FEM job (6900K + 128gb of RAM, SSD), no MVRC.

[rjsa]

Almost right:
you need to change from this:

• application simpleFoam;
  startFrom startTime;

to this:

• application simpleFoam;
  startFrom latestTime;

But how should we get it restarted? what's the command line? Because I've always seen it start from 0 after doing that.

[wb92]

Next year I hope to be back to the brute force of the workstation I have in office.

what kind of specs that workstation has?

An old (2013) Intel i7, with a huge quantity of RAM but a very slow clock (slower then the original clock, I now use it as a backup server and it has temperature limits).

I also have a more modern and powerful workstation in office, but recently I had to dedicate it to a FEM job (6900K + 128gb of RAM, SSD), no MVRC.

so still MVRC_Long would take like whole day on such set, right?
I wonder what config LVDH/Andre has.

------------------------

But how should we get it restarted? what's the command line? Because I've always seen it start from 0 after doing that.

you have to change number of iterations as mentioned here:

Let's say you want to continue calculation after 2000 iterations and run additional 500 iterations (2500 in total):
your controlDict will look like that

application simpleFoam;
startFrom startTime;
startTime 0;
stopAt endTime;
endTime 2000;
deltaT 1;
writeControl timeStep;
writeInterval 2000;
purgeWrite 0;


You need to change it to:

application simpleFoam;
startFrom latestTime; [EDiT: changed according to LVDH note]
startTime 2000;
stopAt endTime;
endTime 2500;
deltaT 1;
writeControl timeStep;
writeInterval 500;
purgeWrite 0;

and then run it as usual, but I noted that MantiumFlow creates more than one controlDict, so you should change these parameters in all of them, just to be sure - @LVDH/Andre - can you tell exactly why there are more than one controlDict?

[CAEdevice]

Ok, all clear. Last question: should I edit the runCase file? If I launch the orignal one it start recreating the mesh. Should I erase the blockMesh and snappyHexMesh lines?