search

IEAWindTask37 / IEA-22-280-RWT

Repository for the IEA 22-MW offshore reference wind turbine developed by the IEA Wind Task 55 REFWIND
Apache License 2.0
49 stars 18 forks source link

OpenFAST Cd0 always zero #73

Closed pelljam closed 2 weeks ago

pelljam commented 3 months ago

Hi,

I just wondered if it was intentional that Cd0 (2D drag coefficient value at 0-lift) is zero for all airfoils?

Thanks,

James

ptrbortolotti commented 3 months ago

Hello James, this is a great catch. It took me a bit of digging to understand why Cd0 is set to 0. I found that back in 2021 I zeroed out this term during a V&V study because things weren't matching, and zeroing Cd0 helped greatly. This happened in WISDEM commit https://github.com/WISDEM/WISDEM/commit/4271c970f13abeef4bd4e82d1c603c7335c915a6 However we now know that we were plagued by multiple other issues in OpenFAST, and so that zeroing of Cd0 might have simply be a lucky coincidence. I will try to turn it back on and see what happens. It will take me a few days to run all the checks, but it's high among my todos

ptrbortolotti commented 3 months ago

I've pushed new airfoil files here https://github.com/IEAWindTask37/IEA-22-280-RWT/tree/openfast_fixes I'm running some tests to make sure they perform as expected

ptrbortolotti commented 3 months ago

Hello again, Once merged, this PR https://github.com/IEAWindTask37/IEA-22-280-RWT/pull/75 will fix the issue of Cd0. Note however that I haven't seen much effect (almost anything) in the time-domain turbulent-wind results. I have not processed the linearized-domain results yet. However generating a good Campbell by linearizing the OpenFAST with unsteady aero airfoil turned on is a little time consuming at the moment, so this will take me a while

pelljam commented 3 months ago

Thanks Pietro. For what it's worth, I observed a noticeable difference between OrcaFlex and OpenFAST, in some time-domain results, which I can attribute to this data (because once I set Cd0 for all airfoils in OrcaFlex, the results converged). Although off the top of my head, I can't remember the details. I can have a closer look, if you are interested? Which UA model are you running? My investigation uses the González model, possibly that choice will dictate the sensitivity to this data?

ptrbortolotti commented 2 months ago

I've been toggling between UAMod 3, which works better with ElastoDyn (@dzalkind can confirm), and UAMod 4, which works better for BeamDyn. We suspect that there is a dependency to the time step size, which must be very small with BeamDyn (we are close to fixing that). The two models are different implementations of the Beddoes-Leishman unsteady aero model https://openfast.readthedocs.io/en/main/source/user/aerodyn/theory_ua.html#beddoes-leishman-4-states-model-uamod-4