Dear NREL community,
I’m trying to detect the frequencies of a NREL 5MW onshore reference wind turbine by post processing the PSDs of blade & tower tip accelerations. The 2 main excitations I implement are white gaussian singal (I implemented that using a uniform-wind type file, where simply the wind velocity varies with each time step) and Kaimal turbulence. My problem is that, when I excite the structure with Kaimal turbulence, the PSDs - especially of the blade- are quite unclear and not all modes are excited. Below I have attached the PSDs for Kaimal and then for white noise excitation.
I exported the Output full-field time-series data of Kaimal turbulence from TurbSim by activating “WrADTWR”. I isolated the wind speed time series on the hub height and plotted its spectorgram in order to see which frequency range it excites. Considering the image below, it seems quite low and I assume that this is the main reason that the PSD is unclear. Is there any way to allow for tuning of the parameters of a Kaimal-type spectrum in order to increase this frequency content? By the way, I tried some other turbulence models (von Karman, Great plains low level jet, even NWTCUP with coherent structures), but I always received more or less the same low frequency content.
The TurbSim parameters I implemented where a 31*31 Grid, a time step of 0.01 sec and Analysis & Usable time of 650 sec. Regarding the Kaimal model parameters, I defined turbulence characteristic of 16.67, normal turbulence type “NTM”, mean wind speed of 12 m/s and a Power Low wind profile type.
I’m a bit surprised by your PSDs for the Kaimal spectrum. I would expect the blade-tip to have strong peaks at 0P, 1P, 2P, 3P, 4P etc. as the blade in the rotating frame of references passes through the turbulent wind field, which I don’t really see from your simulation. And I would expect the tower-top in the fixed-frame of reference to have strong peaks at 0P, 3P, 6P, etc., which I also don’t really see from your simulation. Perhaps your simulation is not run long enough to get statistical significance in the results?
I’m less familiar with how the PSD should look when excited by uniform, but time-varying wind.
Regardless, the natural frequencies of the land-based NREL 5-MW baseline turbine in a parked condition are published in the NREL 5-MW turbine specifications report: nrel.gov/docs/fy09osti/38060.pdf.
I would expect wind spectra to be more energetic at low frequencies. The spectra implemented in TurbSim, as well as example PSD versus frequency plots, are included in the TurbSim User’s Guide: nwtc.nrel.gov/system/files/TurbSim.pdf. You should be able to use this information to understand how the spectra change with variations in the TurbSim input parameters.
I guess the peaks that you describe are better depicted in the following PSD, which by the way refers to steady wind conditions with a mean wind speed of 10 m/s. I believe that more or less the same peaks appear in the Kaimal turbulence PSD, just much less clear and obvious. I did follow your suggestion and carried out an analysis of 3000 sec, which was based on a TurbSim input file of Analysis Time also 3000 sec and a time step of 0.01 sec. The PSD was identical to the one I attached, no improvement was observed.
I have read the NREL 5-MW turbine specifications report. Generally, from the PSDs of various excitations I can identify 2 distinctive frequency peaks of the tower (around 0.312 Hz and 2.9 Hz) and 2 for the blade (around 0.66 Hz and 2.03 Hz). Do we expect that a PSD clearly shows all of the natural frequencies, i.e in FA and SS directions as described in the aforementioned report? Some of them are very closely spaced. I have tried changing the angle of wind excitation to observe if different modes are excited but there was no significant difference.
Regarding the Kaimal-Turbulence PSD: Is it normal that the spectrogram of the turbulent wind excitation shows a frequency content of such a small range? I have read the TurbSim manual but still cannot figure out a way to increase this frequency content just by changing any of the parameters. I don’t know if modifying the source code is necessary. In any case, I would appreciate any suggestions on improving my analysis.
Once again, I appreciate your help.
Your new plot looks clearer, but still not exactly what I would expect. Perhaps that’s because your plotting the y-axis in dB, when I would normally see a PSD plotted in [units]^2/Hz, where “units” is the unit of the time-series variable. Perhaps you’ve also applied some filtering in the calculation of the PSD.
Regardless, it is not necessary easy to identify natural frequencies through PSD analysis unless the source of excitation excites the modes of interest and that response is stronger than the direct excitation.
Appendix G in the TurbSim User’s Guide shows example spectra of the turbulent wind velocity components, where it is clearly seen that energy dissipates at higher frequencies. Equation (21) defines the IEC Kaimal spectra.