I have a question on the unsteady aerodynamic (dynamic stall) model.
Currently, I am running stand-alone AeroDyn test cases to get to know the tool and how it works.
I’ve noticed that activating the UA model causes a drop in Cl and as a result a drop of the predictid AeroPwr.
The BEM part should be fine, becaues the converged AOA and induction coefficients are not affected by switching UA on.
My initial guess was that is should be due to worng dynamic stall parameters (which are obtained from AirfoilPrep, so I played with them and the results don’t seem to be that sensitive to the inputs. I always get the remarkable drop in lift and power in the second time step, please see the attached pictures.
Any hint is highly appreciated.
I would normally expect the unsteady airfoil aerodynamics (UA) model of FAST / AeroDyn influence the variations of lift about the mean, but not to directly influence the mean lift itself. Your results show an unexpected impact of UA on the mean lift.
Which version of FAST / OpenFAST and which UA model of AeroDyn are you using? What UA parameters have been set?
thanks for your reply. I was not anticipating a change in mean values either.
My results are from AeroDyn_v15.04.02 (standalone mode).
I am using the Beddoes-Leishman unsteady model with Gonzalez’s method (UAMod = 2)
The UA parametes are the following:
They originate from the AirfoilPrep excel-tool.
Although filtCutOff = 200 Hz seems quite high, I don’t see any obvious errors in your other input. Do you get improved results by setting filtCutOff = “DEFAULT” (20 Hz), or lower? Do you have similar issues when using other UAMod options, e.g. UAMod = 3?
thanks for your hints and for checking my input.
I’ve tried UAMod=3 and also reducing filtCutOff down to 2 Hz. But I am still getting the same results.
I was checking the AeroDyn 5MW benchmark, where I noticed that the same issue happens there as well.
Please have a look at the figure below. It compares the results with UA being switched On and Off.
They are generated from the default AeroDyne example: “NRELOffshrBsline5MW_Onshore_AeroDyn15_Test01.dat”
I’ve not made any changes in the input files rather than activating/deactivating UA.
Activating UA causes a remarkable drop in RtAeroPwr and the lift coeffienct in two blade sections out of the plotted four.
Perhaps the UA hysteresis loop is not symmetrically balanced about the static lift curve for this airfoil and UA settings, such that adding UA changes the mean lift? Have you tried plotting the UA hysteresis loop and the static lift curve versus angle of attack on the same plot?
It may help to use the UA driver to understand this behaviour. If you are unaware, there is a yet-to-be-documented standalone driver for running only the UA submodel of AeroDyn v15 in isolation that is included in the AeroDyn v15 and OpenFAST archives. You may want to play around with this to better understand the influence of UA.
it is an usymmetric hysteresis loop in deed. Here the comparison between static lift and UA lift for B1N4 and B1N5:
I am a newbie in UA and should definitely dig deeper into its theory. Do you think it makes sense (physically) for the unsteady lift coefficient to drift away from the static one? Or can it be due to not prefectly set UA parameters in the airfoil input file?
I’ll try the standalone UA driver too, thanks for the hint.
Well, that explains why enabling UA shifts the mean lift. I would expect inconsistencies between the UA parameters and the static airfoil data to be the cause. Are you still using unmodified airfoil data from the AeroDyn v15.04.02 CertTest? For example, are you using unmodified NRELOffshrBsline5MW_Onshore_Test01.dvr, NRELOffshrBsline5MW_Onshore_AeroDyn15_Test01.dat, NRELOffshrBsline5MW_AeroDyn_blade.dat, and DU25_A17_10Hz.dat?
from the files you’ve mentioned, I’va just playes with ‘NRELOffshrBsline5MW_Onshore_AeroDyn15_Test01.dat’ to activate/deactivate UA. The rest are left untouched.
OK, thanks for confirming. I’ve now posted this problem as an issue in the OpenFAST r-test repository: github.com/OpenFAST/r-test/issues/20. Hopefully this will start the process to get this bug in the airfoil data input file(s) fixed. Please feel free to add further comments there. And let us know if you correct the airfoil data input file(s) yourself.
We looked into this issue a bit and believe we understand why there is an offset between the static airfoil data and the hysteresis loop with UA enabled. The issue is related to nonlinearity in what should be the linear region of the lift curve. When the lift curve is linear, the issue goes away, but if there is nonlinearity in that region, a shift is introduced where the UA solution jumps down to the linear region. See the results below from the IEA Wind Task 30 OC6 Phase III project.
One solution is to ensure that the linear region of the lift curve is, in fact, linear. The other solution is to switch from using UAMod = 3, to the new UAMod options recently introduced in AeroDyn, e.g., UAMod = 4 (HGM) and UAMod = 5 (HGM + one additional state): github.com/OpenFAST/openfast/pull/729, which don’t suffer the same problem.
Realistic lift curves are highly non-linear. Its very hard to match linear numeric modeling to non-linear experimental data, except in the flat spots-