Dear Jason,
in preparation for some structural analysis of the 5 MW reference wind turbine I simulated a shutdown. For this I used a steady wind with v=12 m/s.
I my simulation for this load case I used ElastoDyn and BeamDyn. To compare these two tools I looked for the torque at the blade root, the tip deflection and the sectional forces along the relative blade at the x,y,z direction
It seems that there is an offset at every force, torque or deflection I investigated. I used the reference turbine. Therefore, there shouldn’t be any mistake in the model itself. If I compare it to your investigation in the article [1] my results don’t have a good agreement. I don’t know what my mistake is.
Do you know, what the cause for this offset is?
BeamDyn - Simulation
time increment 1E-3
QuasiStaticInit - True
Rhoinf - 0
Quadrature –
Trapezoidal
Order of Element 8
Figures Shows
Deflection: Comparison between TipDxC1 - B1TipTDxr and so on
Torque: Comparison between RootMxb1 - B1RootMxr and so on
Section Force: Comparison between SpnFxb1 - B1NFxl and so on
Best regards,
Moritz Pick
[1] Qi Wang, Michael A. Spraque, Jason Jonkman, Nick Johnson and Bonnie Jonkmann: „BeamDyn: a high-fidelity wind turbine blade solver in the FAST modular framework”, Wind Energy (2017)
Dear Moritz,
Which version of FAST / OpenFAST are you using? There have been many bug fixes made to BeamDyn over the past couple of years (including to the sectional load outputs). When running BeamDyn, I would suggest using the newest version of OpenFAST (dev branch).
When comparing the tip deflections, the “xc” axis in ElastoDyn is not the same as the “xr” axis in BeamDyn (if there is a nonzero pitch angle). Instead, I would compare the deflections using the “xb” axis in ElastoDyn against the “xr” axis in BeamDyn.
Best regards,
Dear Jason,
I’m using the latest OpenFast version. The comparison is between TipDxb1 - B1TipTDxr, it was a typo. Unfortunately the the pictures of deflection and spanload got mixed up. The investigation of the spanload are at 350 s of the simulation time.
best regards
Moritz Pick
Dear Moritz,
From the second plot of the RootMx, it looks like the two simulations are at different azimuth angle at 350 seconds. In that case we would expect the blade spanwise loading to be different. Do you see differences in the spanwise loading if you investigate at a time when the azimuths align (maybe around 370 seconds)?
Regards,
Andy
Dear Jason,
you are right, the Azimuth is different. Apparently there is a difference in the pitch wich results in a different Azimuth and of course force. But theses simulations have the same initial parameter. Is there any chance that the pitch controll don’t work as it should?
best regards
Moritz Pick
Dear Moritz,
Differences in the blade deflection between ElastoDyn and BeamDyn often result in differences in the blade-pitch angle. For example, if the blade twists in BeamDyn (which it can’t do in ElastoDyn because ElastoDyn has no blade-torsion degree of freedom), the aerodynamic torque will differ and the pitch angles will be different to compensate for the different torques.
You say that both simulations are initialized withe the same initial parameters, but you are only showing results after 300 s. My guess is there is a transient in the blade-pitch response in these results between 0 and 300 s.
Best regards,
Dear Jason,
I ran the the Simulation from 0 to 300 s, with a wind velocity of V_hub 24 m/s . Both Simulation are different compared to the wind-speed-relation of 24 m/s and a pitch-angle off 22.35 deg.
best regards
Moritz
Dear Moritz,
I think this makes sense. The simulation is close to rated power which I expect will produce the largest torsional moment along the blade. ElastoDyn does not have any torsional degree of freedom, so the tip does not twist relative to the root. Therefore in the ElastoDyn simulation, the controller then pitches the blade to 22.35 degrees towards feather to hit the rated power.
With the BeamDyn simulation, there is a torsional degree of freedom along the blade span. As a result the tip of the blade twists relative to the root (twists away from power to feather). This means that part of the blade is already pitched towards feather due to the aerodynamically induced moments, so the controller does not have to pitch the blade as far as in the ElastoDyn case to achieve the same overall effect.
The difference between the ElastoDyn and BeamDyn case is slightly larger than I would have expected (I would have expected 2 to 3 degrees). However this may be reasonable depending on cone angle, shaft tilt, and unsteady aerodynamics used. Perhaps @Jason can comment on this.
Regards,
Andy
Dear Andy,
thank you for your answer.
Best regards
Moritz
Hi Moritz,
I ran a quick check of the NREL 5MW model at 24 m/s wind speed. For this check, the only change from the standard regression tests was the wind (current dev branch, git hash a51fffaa). I set the wind to 24 m/s at hub height with 0.2 for the power law exponent. The cone angle, shaft tilt, and all other settings are identical to the standard test cases.
What I noticed is I get a smaller difference between the blade pitch in BeamDyn (20.7 degrees) and the ElastoDyn simulation (21.71 degrees). Did you change any turbine parameters (shaft tilt, cone angle, etc), or use a different wind profile than I did? If so, that would explain why you see a larger difference in blade pitch than I do.
Regards,
Andy
Dear Andy,
I didn’t changed any parameter and use a steady windprofile. Therefore I do not know where the deviation comes from.
regards
Moritz
Dear All
thank you for your last answers. I have another question about BeamDyn and ElastoDyn.
I compared the sectional loads of ElastoDyn and BeamDyn with an analytical calculation. Therefore, I made a simple simulation with openFAST 2.2 (dev) and the 5MW reference turbine. The wind is steady at 11.4 m/s.
AeroDyn:
• WakeMode 1
• AfAeroMod 1
ElastoDyn:
• WakeMode 1
• AfAeroMod 1
• All Degree of Freedom are “false”
BeamDyn:
• quadrature trapezoidal
• QuasiStaticInit true
• Element order 8 (9 Nodes)
• The Other parameter are on “default”
ServoDyn:
PCMode 5
The loads of ElastoDyn are
“Spn1MLxb1, Spn1MLyb1, Spn1MLzb1, Spn1FLxb1, Spn1FLyb1, Spn1FLzb1”
There are 19 nodes and the output nodes are 1,3,5,7,10,13,15,17, 19
The loads of BeamDyn are
“N1Fxl, N1Fyl, N1Fzl, N1Mxl, N1Myl, N1Mzl”
The loads shown have the same Azimuth angle (90°).
The reactions at the blade tip are different and not what I expected. The loads of BeamDyn are too high, I think. The force and moments should be close to zero. The main difference should be a result of the different structure model. As you can see at the following figures the investigated aerodynamic loads are as expected.
At the root section the moments are almost the same except the torsion “Mz”.
Could you can explain the reason for the deviation?
here is aerodynamic Pitchmoment
And here are the moments at the root
best regads
Moritz
