search

Exawind / hfm-fsi

Input files and data for HFM FSI Work
0 stars 0 forks source link

Run NREL-5MW full power curve with FSI #64

Open psakievich opened 5 months ago

psakievich commented 5 months ago

Started the implementation of input files https://github.com/Exawind/exawind-cases/pull/18

kevmoor commented 5 months ago

Phil has it 50-60% set up, collecting data. All of the scripting process was abstract able, just need updated openfast inputs for running with dev. Also need other updated inputs like controller etc.. Have draft pull request in exawind-cases as opposed to Ganesh's repo. Meshes are in /projects/hfm/ndevelder/meshes

Need meeting or slack regarding what the inputs should be. Andy Platt may be able to give the needed files. Files may be in rtest as well, just have to minor change 6 or 7 files specific to the FSI runs.

kevmoor commented 5 months ago

Existing files show documentation of what changes are needed via git diff, but need to identify new things from the latest openfast. Boiling down what needs to be done could be done as a post step and should be straightforward. Current openfast model, take files we need and drop in the FSI directory and diff the changes and modify/revert as needed, test, iterate.

Phil pull rtest files, Need someone to look into process csv file and potentially change python script, Nate to take on.

pscrozi commented 5 months ago

We can run the power curve now, but we're getting stopped by the build issues we're seeing on eagle. Phil has a message out to Ganesh about taking over his builds that were working ok. We might decide to try another fresh build as well. Might experiment with a newer version of MPT (HP's version of MPI) on eagle. Consider reaching out to sysadmins about MPI issues others might be seeing.

Could consider running on amber, but it is busier now and still getting segfaults. Probably best to stick with eagle.

psakievich commented 5 months ago

Kevin was able to get his build going today. There was an issue with the openfastcpp executable in the wrong place. Phil has also created a build and was able to test some regression tests across multiple nodes. He will run a turbine, and if successful, publish a module for the team to use.

kevmoor commented 5 months ago

Ganesh: Hey guys.. I’m trying out the NREL 5MW.. I haven’t been able to progress beyond running the openfast precursors yet I’m waiting for jobs to try out 1 FSI case right now

Eagle queue seems to be pretty full right now.

Neil: may have found issue with split mesh having tilt already in it, then FSI moves it by tilt again. Hoping to take out tilt transformation in NALU and that should fix it. Double tilting causes the sim to fail since meshes not on top of each other

pscrozi commented 5 months ago

Two issues: (1) mesh; (2) build. Nate thinks it is all the mesh that is causing the problem.

pscrozi commented 5 months ago

Nate has a run going that may confirm what Ganesh was seeing in the mapping yesterday with the negative volumes.

pscrozi commented 5 months ago

All cases ran out to 10 revs and looking good. Going to do restarts now. Targeting 20 revs.

psakievich commented 5 months ago

Still working on setting up the restarts.

pscrozi commented 4 months ago

Ganesh has 10 cases running. 2 highest wind speeds are waiting in the queue. 2 wind speeds in Region 3 have gone past 20 revs. Trying to get to 35 revs. Should be done by Monday.

1 crashed on eagle due to MPI errors.

kevmoor commented 4 months ago

Ganesh ran 5, 7, 8, 9, 10, 12, 14, 17, 20.

23, 25 m/s have run 47k/50.4k time steps

Will ask if we have enough content for milestone completion next time.

kevmoor commented 4 months ago

From Tues meeting: Ganesh post process 5MW and get curves/power curve

pscrozi commented 4 months ago

All have run. In Region 2, the power is lower than in OpenFAST standalone. Rigid blade simulations matched the power curve spot on. So it isn't necessarily wrong. We should plot them and complete this.

Ganesh has a handle on the differences.