Comparison of FAST and Flex5

Wind & Water Computer-Aided Engineering Software Tools
1

Dear all,

I just compared the results of a 2.5MW turbine, simulated in FAST and in Flex5 by looking at side-to-side acceleration of the tower top. If I do a FFT of that signal I see quite large differences between the two simulations:

  • Amplitudes differ by a factor of 4.
  • Eigenfrequencies are quite similar.
  • Above 3p there is almost no frequency content in FAST, Flex5 however shows peaks at 2nd tower eigenfreqency, 2nd flapwise frequency and in between and beyond.

Of course I understand, that there are differences between FAST and Flex5. But I really wonder why I can see almost no exication of the 2nd tower eigenfrequency and above in FAST.

Parameters of the model are almost identical, it uses the same controller, similar wind files. Has anybody some experience and can tell me why there are such huge differences between the two simulators?

Best regards,
Felix

2

Dear Felix,

I have not run FLEX5 myself, but we have done many code-to-code comparisons (including to FLEX5) and the differences you are reporting are surprising. Please clarify the following:

*If there is absolutely no excitation in the 2nd bending mode frequencies of the blades and tower in your FAST results, are these DOFs enabled in the model?
*The side-to-side damping is very limited from aerodynamics, so, structural damping is likely a large contribution to the overall side-to-side damping. Are the side-to-side structural damping levels specified realistically and identically in both models? It may help to compare the damping levels derived from a linear eigensolution for the modes of interest.
*You only refer to problems with the side-to-side comparison. How do other parameters compare?
*The side-to-side excitation is dictated by the torque and asymmetric loads. How does torque compare, for example?

Best regards,

3

Dear Jason,

thanks for your suggestions.

  • The 2nd bending mode is enabled in the model
  • The structural damping was set to 3.5% for both models. Also the damping in Flex5 was increased to 5%, but there are still higher amplitudes compared to FAST
  • At low frequencies, the results compare well, but FAST lacks the higher frequencies. The attached plot shows the Rotor loads at the LSSTip. First row shows Fx, Fy, Fz, second row Mx, My, Mz. FAST is red.

I really hope anyone has a good idea, what is going wrong here.

Best regards,
Felix

FAST_Flex5_LSSTip.png
FAST_Flex5_LSSTip.png1380×1020 20.1 KB

4

Dear Felix,

Thanks for sharing some of your results,

While I don’t know what is causing the differences, differences in the magnitude of the power-spectral densities can be related to differences in the excitation loads or differences in the inherent damping levels. This suggests two ways to look further into what is leading to the differences:

*To assess possible differnces in the excitation loads, I suggest you first simplify both models by disabling all structural DOFs (simply calculate the aerodynamic loads on the operating, but rigid, turbine). Are the aerodynamic load levels equal across all frequencies in this case? One difference could be related to the wind input. For example, my understanding is that FLEX5 uses polar turbulence grids whereas FAST with AeroDyn use rectangular grids. This tells me that you are likely not using the same wind input in both simulations (unless you modified the source code of one of the models)–so, how do the spectral contents of the specified wind compare? Also, if you interpolated one wind field to make the other, then the other would have a smoother wind condition and thus lower variation in excitation.

*Each code may interpret the specified modal damping differently. For example, the modal damping of the tower in FAST is specified as the modal damping of an isoloted tower–this will be different than the damping of the tower bending mode in the full system that contains couplings to the rotor-nacelle assembly. It would help to compare the damping levels derived from an eigensolution of a linearized full system model. In FAST, you can derive this through the linearization functionality (AnalMode = 2).

I hope that helps. Please let us know what you learn.

Best regards,

5

Dear Jason,

part of the problem seems to be, that the damping ratio for the blades and the tower where copied from the 1.5MW-model supplied with fast. These damping ratios are about 3.5% and above, although the FAST manual suggests to use values between 0.5% and 1.5%. The Flex5 model used much smaller damping values. What is the reason that the model supplied with FAST uses such large damping values?

I will go through the comparison of FAST and Flex5 in more detail on monday.

Best regards,
Felix

6

Dear Felix,

I agree that those damping values from the WindPACT tower and blade model are high. They were derived to give the same damping in FAST as what was modeled in an earlier ADAMS model. This ADAMS model was made outside of NREL and I’m not sure how the original damping levels were derived. More reasonable structural damping values are 0.5-1.5% for steel towers and 2-3% for composite blades.

Best regards,