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

[wb92]

ah, I forgot there's runCase file which stores all OpenFOAM operations, including meshing. Yes, in this case you need to comment out/delete everything what's related to meshing, so it would start from:

• mpirun -np $nCPUs simpleFoam

You don't need to run potentialFoam again

maybe suggestion for the new season to split scripts for meshing/calculations/post-pro ?

[CAEdevice]

ah, I forgot there's runCase file which stores all OpenFOAM operations, including meshing. Yes, in this case you need to comment out/delete everything what's related to meshing, so it would start from:

• mpirun -np $nCPUs simpleFoam

You don't need to run potentialFoam again

maybe suggestion for the new season to split scripts for meshing/calculations/post-pro ?

Thanks for the confirmation,I am going to try tomorrow.

[yinlad]

Hmmm I had a 'fast' run going for 12 hours and it had only got to 'time' 225. Took 2 hours just to build the mesh. I fear I may have done something wrong on that front. I used the demo mantiumflow for that run. Having received a log in for the challenge site earlier today, I downloaded the version of mantiumflow from there and tried to build/run from scratch but just get a 'cant run runFile' error when trying to run a build. But am able to run that same build in the demo version (but haven't left it to try and resolve just yet).

Any of you guys on a discord server or something of the like? Hate to be that guy but cant help but feel I'm doing something wrong. Or perhaps my workstation isn't up to the task. In any case I feel this thread isnt really the right place to be asking for/receiving tech support haha

[CAEdevice]

2 hours to build mesh is reasonable , 12h to reach iteration 225 not. Which hardware are you running on?

[yinlad]

This machine is just a 6 core Ryzen 5 with 16gb ram. Never had issues with CPU maxing out during any of it, but RAM seemed the limiting factor during the openfoam part, once the mesh had built.

I think I need to be looking at more like 64gb of RAM minimum, right?

EDIT: I also changed the mantiumflow settings to run on 6 cores, up from the default 4. Perhaps there's optimisation issues when changing from the default 4 core settings?

[CAEdevice]

The hardware seems ok! 16gb are enough, also 8gb are ok for mvrc_fast and most of the simulations done with MVRC template.

Can you see how many elements did the mesher create? Are you sure the model has the right scale? Could you share the mesh.log file and any log you can find in the case folder?

In the first steps of mesh creation, the mesh log reports how large is the bounding box in meters.

You should have less than 10x10^6 cells (usually around 5x10^5) with the standard MVRC template.

[yinlad]

I've since deleted it all while trying to get it to start again I figured I'd regret that. If it's any help I was basically running machin's car from viewtopic.php?f=14&t=28435 this thread. But with a basic front wing I put together. Perhaps my front wing is the issue.

I'll see about trying to run his car as-is tomorrow, see how long it takes.

[machin]

I was basically running machin's car from viewtopic.php?f=14&t=28435 this thread

Check the scale on that model... FreeCAD outputs the model 1000 times bigger than it needs to be, so I reduced that down in paraview before making the submissions to Andre..... could this be the problem?

[yinlad]

I was basically running machin's car from viewtopic.php?f=14&t=28435 this thread

Check the scale on that model... FreeCAD outputs the model 1000 times bigger than it needs to be, so I reduced that down in paraview before making the submissions to Andre..... could this be the problem?

If I measure the stl in freecad its 1.97mm across the front wing and 1970mm in the actual freecad file. Which I believe is correct for running in mflow.

I'll take a look at paraview though. To scale it down I was using multiple scaled down bodies using draft clone, which may have introduced or retained more complexity than necessary

[rjsa]

Hi Matteo, did it work?

[CAEdevice]

Hi Matteo, did it work?

I must have done something wrong... it starts from iteration 0!

[rjsa]

Hi Matteo, did it work?

I must have done something wrong... it starts from iteration 0!

That's what's happening here.

[LVDH]

ah, I forgot there's runCase file which stores all OpenFOAM operations, including meshing. Yes, in this case you need to comment out/delete everything what's related to meshing, so it would start from:

• mpirun -np $nCPUs simpleFoam

You don't need to run potentialFoam again

maybe suggestion for the new season to split scripts for meshing/calculations/post-pro ?

Yes, you need to run everything after the potentialFoam line. If have now executed the entire script once, you possibly have overwritten your previous iterations. You can check what the latest saved timestep is by checking a processor* folder. If you sill find a folder called 500 in there and you switched the startTime in the controlDict to latestTime, it should work.

I was basically running machin's car from viewtopic.php?f=14&t=28435 this thread

Check the scale on that model... FreeCAD outputs the model 1000 times bigger than it needs to be, so I reduced that down in paraview before making the submissions to Andre..... could this be the problem?

If I measure the stl in freecad its 1.97mm across the front wing and 1970mm in the actual freecad file. Which I believe is correct for running in mflow.

I'll take a look at paraview though. To scale it down I was using multiple scaled down bodies using draft clone, which may have introduced or retained more complexity than necessary

Having the scale in the stl wrong could indeed be a problem, explaining this. However MFlow should actually report if your geometries are too big or too small. Paraview should show you that your stl file of the wing is about 2m wide.
Often a problem is that the geometries are open and mesh leaks inside which causes the mesh to be too big and your system to run out of memory.
You can send me your log files or even the entire case if you want so I can have a look.

[yinlad]

I have worked on a few parts tonight and scaled them down using paraview instead of the garbage way I did it before. The parts worked out a lot better using that method.

I'm just kicking off a new MFlow run overnight to see how far it gets by the time I wake up. I changed mflow back to 4 cores in case 6 was causing an issue for some reason. Christian also DMd me and mentioned that sometimes a run can bug out and the time intervals end up taking about 10x longer than they should, so perhaps I ran into that on my first run attempt.

Also thanks all for being so helpful, it's really appreciated! Soon I'll be able to confirm my suspicions that I have no idea about aero design

[Koldskaal]

Often a problem is that the geometries are open and mesh leaks inside which causes the mesh to be too big and your system to run out of memory.

I think this must be what happens to me sometimes. I have many parts that overlap each other, could this confuse the mesher?
I usually just use the stls as exported by solidworks. So they shouldnt be open, but who knows.