Tower Eigenfrequencies of NREL 5MW Turbine

Dear Dr. Jason.

When I used the FAST linearization, I obtained following results regarding tower top mass properties. The first column is given in NREL, and the second column is values calculated by me using FAST linearization (v7.00)

Title 1st column 2nd column
Topmass 3.50000E+05 3.50000E+05
cm_loc -4.13775E-01 -4.13774E-01 (From Tower center line)
cm_axi 1.96699E+00 1.96699E+00 (From Tower Top)
ixx_tip 4.37000E+07 3.86484E+07 (w.r.t. CM of RNA)
iyy_tip 2.35300E+07 2.35557E+07
izz_tip 2.54200E+07 2.54085E+07

In order to calculate this, I used very small tower and platform mass and located platform reference point* at 89.56699 from MWL (z coordinate CM of RNA). Also, I used default angle of azimuth value and initial azimuth. (0.0 AzimB1Up & 0.0 Azimuth ) . When FAST linearization is used, smaller Ixx_tip is obtained than when Adams is used. I don’t know why this difference is generated. I am not sure whether Ixx = 3.86484 is correct or I make some mistakes in linearization input files. Could you check this? I used NRELOffshrBsline5MW_OC3Hywind.zip in wind.nrel.gov/public/jjonkman/NR … Bsline5MW/ I changed some values like followings for linearization.

In NRELOffshrBsline5MW_Floating_OC3Hywind.fst
2 AnalMode
0 PCMode
0.0 RotSpeed

In NRELOffshrBsline5MW_Linear.dat
False CalcStdy
2 MdlOrder

In NRELOffshrBsline5MW_Platform_OC3Hywind.dat
0.0 TwrDraft
0.0 PtfmCM
-8.9566990E+01 PtfmRef
0.0 PtfmMass
0.0 PtfmRIner
0.0 PtfmPIner
0.0 PtfmYIner
0 PtfmLdMod

In NRELOffshrBsline5MW_Tower_OC3Hywind.dat
Very small mass density 0.0000001kg/m

  • If PtfmRef is less than TwrDraft, an error is occured, but I modified the source code so that the error is ignored.

Sincerely,
Daniel Kim.

Dear Daniel,

I’m not sure. I’m quite confident that ixx_tip = 4.370E7 kg*m^2 is correct for the NREL 5-MW turbine.

Perhaps there is a problem because you overwrote the error that PfmRef < TwrDraft? Do you get the correct result if you set PtfmRef = 0.0, linearize to obtain the mass matrix, then transform the inertias from the tower base to the rotor CM?

Best regards,

Dear Dr. Jason.

Even though I calculated Ixx w.r.t the tower base (0.0 TwrDraft, 0.0 PtfmCM , 0.0 PtfmRef) without modifying source code and transformed Ixx w.r.t CM of RNA, the result was the same as 3.8648381E+07.

Ixx w.r.t. tower base: 2.8464364E+09
Ixx w.r.t CM of RNA: 2.8464364E+09 - 35,000*89.566990^2 = 3.8648381E+07

I used the NRELOffshrBsline5MW_OC3Hywind.zip in wind.nrel.gov/public/jjonkman/NR … Bsline5MW/ , and I didn’t change other variances except for values that I listed above. I have to make the consistent input files regarding tower-top between BModes and FAST in other case, so I am first checking the existing NREL 5MW wind turbine, but the results are not matched to the given values, which is calculated by ADAMS. How can I confirm it? Have you obtained inertia results of NREL 5MW tower-top by using FAST linearization and compared them to the results obtained from ADAMS?

Sincerely,
HyoungChul Kim.

Dear HyoungChul,

Good question. I was able to reproduce your results using a linearization analysis with FAST v7. While I have the old ADAMS models, unfortunately, I no longer have a working MSC.ADAMS license to check the aggregate mass results we’ve used in the past. There must be something set differently between these models, but I don’t recall seeing these differences before. I’m now just as confused as you.

Best regards,

Dear Jason,

I have a question about the izx_tip (cross product of inertia about z and x reference axes) value of the BModes v1.03.01 Input File.
When I calculate the first 10 modes of the NREL 5MW Tower, there is no influence of the value on the BModes results. The modes and the frequencies are exactly the same. I tested ixz_tip with a value of 0, 1.169E6 and 1.169E7.

So, when is it necessary to enter the izx_tip value?

Attached the results of BModes in the command window.

with ixz_tip = 0:
      eigenvalue(450) =  0.390794D+11        mode  1 frequency =      0.329475
      eigenvalue(449) =  0.398465D+11        mode  2 frequency =      0.332693
      eigenvalue(448) =  0.126954D+13        mode  3 frequency =      1.877897
      eigenvalue(447) =  0.187182D+13        mode  4 frequency =      2.280244
      eigenvalue(446) =  0.774666D+13        mode  5 frequency =      4.638803
      eigenvalue(445) =  0.921857D+13        mode  6 frequency =      5.060349
      eigenvalue(444) =  0.459755D+14        mode  7 frequency =     11.300868
      eigenvalue(443) =  0.471883D+14        mode  8 frequency =     11.448961
      eigenvalue(442) =  0.169404D+15        mode  9 frequency =     21.692580
      eigenvalue(441) =  0.170534D+15        mode 10 frequency =     21.764779

with ixz_tip = 1.169E6:
      eigenvalue(450) =  0.390794D+11        mode  1 frequency =      0.329475
      eigenvalue(449) =  0.398465D+11        mode  2 frequency =      0.332693
      eigenvalue(448) =  0.126954D+13        mode  3 frequency =      1.877897
      eigenvalue(447) =  0.187182D+13        mode  4 frequency =      2.280244
      eigenvalue(446) =  0.774666D+13        mode  5 frequency =      4.638803
      eigenvalue(445) =  0.921857D+13        mode  6 frequency =      5.060349
      eigenvalue(444) =  0.459755D+14        mode  7 frequency =     11.300868
      eigenvalue(443) =  0.471883D+14        mode  8 frequency =     11.448961
      eigenvalue(442) =  0.169404D+15        mode  9 frequency =     21.692580
      eigenvalue(441) =  0.170534D+15        mode 10 frequency =     21.764779

Thank you for your answer.

Best regards,
René

Dear René,

I’m a bit surprised izx_tip has no effect on the eignensolution, although I haven’t tried your test before. A nonzero izx_tip (when the other cross moments of inertia are zero) implies that the principle axes of inertia are rotated about the y axis. This would lead me to expect that a nonzero izx_tip would effect modes that involve rotation of the tower-top about the x and z axes, unless izx_tip is small relative to the ixx_tip and izz_tip.

Best regards,

Dear Jason,

Thank you for your answer.
After your reply I have tested when izx_tip has an influence on the results. With realistic data for tor_stff (from Table 6-1 of the NREL 5 MW Definition document) instead of a “very large number” (I have 1.0E+60) in the sec_props_file you can see an influence.
Can your recommend me a “very large number” for tor_stff and axial_stff for mimic the FAST model correctly?

Thank you again for your support.

Best regards,
René

Dear René,

To make tor_stff a “very large number”, I would recommend setting tor_stff to be a couple orders of magnitude larger than the corresponding bending stiffness. You can use the same values for the axial_stff (despite the different units). Setting these stiffness values too large may lead to numerical problems, and thus inaccurate solutions, in BModes.

Best regards,

Dear mr Jonkman,
I am trying to ran the exact simulation of the 5MW OWT with distributed springs (OC3 Phase II), in order to check the accuracy of my recompiled version of FAST. My question refers to BModes and the eigenfrequencies. My results are the following:
freq 1: 0.2448 (s-s)
freq 2: 0.2464 (f-a)
freq 3: 1.278 (s-s)
freq 4: 1.357 (s-s)
freq 5: 1.511 (f-a)

The first 2 are pretty close to the right ones. I got a little confused that there were two consequtive side to side frequencies following, but as I read in this post, the 3rd is the tortional (is that right?). However, the 2nd s-s frequency (freq 4) differs from the correct value, i.e. 1.5459. I am attaching both the input files and the output, in case you could kindly spot some mistake.
Thank you in advance for your help.

Best regards,
Christos Lathourakis
OC3_DS_check.zip (17.2 KB)

Dear Christos,

Your BModes input files look fine with me.

Yes, I would guess that mode 3 is the first torsion mode and mode 4 is the second side-to-side bending mode. But you could verify that further and better isolate the tower side-to-side bending modes by increasing the torsional stiffness of the tower in BModes (by an order of magnitude of two).

Regardless, the second tower-bending modes tend to have a lot of coupling with blade bending, which of course cannot be captured by BModes. You’ll only see this effect in your FAST simulation i.e. I’d expect you’d get different frequencies from FAST for the second tower-bending mode than you would in BModes.

Best regards,

Dear mr Jonkman,
thank you very much for your time and your precious advice. Your answers were of great help for me.

Best wishes,
Christos Lathourakis

Dear Jason,

I’m using BModes calculating model of the Tower , but I want to calculate the higher order model in the FAST (more than the second order), so I got the BModes calculation results, and calculate the six order model, then import to ModeShapePolyFitting. XLSX was calculated in the Fore and aft side - to - side six modal coefficient, and verified, addition of each set of coefficients is 1, and I have modified the ElstoDyn - Tower, the corresponding coefficient of the input file you can see in the attachment.NRELOffshrBsline5MW_OC3Monopile_ElastoDyn_Tower.rar (1.71 KB)
Then I ran FAST, but there was an error, and the error message is shown below.
NRELOffshrBsline5MW_OC3Monopile_ElastoDyn_Tower.rar (1.71 KB)

What I want to ask is whether the higher-order (more than second-order) modes of Fore-Aft and side-to-side can be calculated in FAST, whether they can be implemented in FAST and, if so, how should I modify them.

Looking forward to your reply, thank you

Best regards
Zhou.Lin
1544691082(1).png

Dear Zhou.Lin,

The archives you attached appear to be corrupted, so, I can’t review what you attached. But regarding the invalid numeric input error, as with any input file format problem, I suggest that you use the Echo option from the input file to debug errors in the input-file processing.

I’m not sure what you mean be “higher order”. I assume you are referring to bending modes of the tower higher than the first and second modes–is that correct? Without customization of the FAST source code, the structural model of FAST v7 and the ElastoDyn module of FAST v8/OpenFAST are limited to two fore-aft and two side-to-side bending-mode DOFs in the tower. If you’ve looked in the source code, you’ve likely noticed that most things are parameterized into loops through enabled DOFs. As such, it would not be too difficult to customize ElastoDyn to allow for additional bending-mode DOFs in the tower. However, we’ve not done this ourselves because as you introduce higher-frequency modes/DOFs, likely modes other than bending will become important (e.g., torsion). To capture effects other than bending requires a change to the beam formulation used for the tower. Another option would be to model the tower with the SubDyn module instead of ElastoDyn. SubDyn allows for any number of modes and includes torsion (and other effects), but SubDyn is purely a linear formulation whereas ElastoDyn captures some geometric nonlinearities, which may be important for a tubular tower.

Best regards,

Dear Jason,
Thank for your reply,I see what you mean. Thank you for your kindness.

Best regards,
Zhou.Lin

Dear Dr.Jonkman,
Could you please send me “ModeShapePloyfitting.xls”. Thank you very much.

My email: luyun86674838@163.com

Dear Yun.Lu,

The spreadsheet ModeShapePolyFitting.xls is included in the FAST v8 archive (nwtc.nrel.gov/FAST8) in the Utilities directory.

Best regards,

Dear Jason & Dear All,

I am working on the MIT NREL TLP model, which is readily available in Model #23 of FAST. As we know that pitch response and tower bending response of the structure is coupled. I want to obtain the pure pitch frequency of the structure. To get that, I switch TwFADOF1, TwFADOF2, TwSSDOF1, TwSSDOF2 input flags to “FALSE” in ElastoDYn module input file. In this way I obtain a pitch response frequency of 3.48 seconds which is lower than the reference value, 4.52 sec (Matha, 2009 Table 6). When I check the tower displacements, TTDspFA, response is zero with respect to tower base coordinate system.

1 st question: Does this value (T=3.48 seconds) makes sense for the pure pitch period? I cant see any reference for that.

Alternatively, rather than disabling the tower flexibility from input flags TwFADOF1 and TwSSDOF1, I increase the tower stiffness value with increasing the tower adjustment factors from ElastoDyn_Tower.dat. I can increase the stiffness up to 5 and see the smaller tower displacement response with respect to the original structure.
Further increase results with this WARNING: HydroDyn_CalcOutput: Angles in GetSmllRotAngs() are larger than 0.4 radians.
Following that MAP aborts the simulation with the following message “Line Failed.”

Alternatively, I also tried to change DISTRIBUTED TOWER PROPERTIES. I’ve increased the TwFAStif & TwSSStif 100 times throughout the tower but ElastoDyn reports “Small angle assumption violated in SUBROUTINE SmllRotTrans() due to a large tower deflection” and the solution aborts again.

2nd question: How can I accomplish a rigid tower and conclude the analysis in the second approach?

Any ideas are welcomed.

Sincerely,
Mustafa Vardaroglu

Dear Mustafa,

Yes, disabling the tower DOFs (or increasing the stiffness of the tower) will increase the platform-pitch natural frequency (reduce the period) from 0.221 Hz (1.39 rad/s) to 0.287 Hz (1.81 rad/s). This can be seen in the RAO plots for the 27-m spoke in (Matha 2009, Figure 26) from WAMIT, which assumes a rigid tower.

Increasing the stiffness of the tower will increase its natural frequency, which must be resolved by the time-integration scheme, and will thus require a smaller time step, as has been discussed many times on this forum.

Best regards,

Hello Jason,
Is there a plan to include tower torsion?

Thanks,
Ian Prowell

Hi Ian,

We’d like to add a tower torsion degree of freedom (DOF) to the ElastoDyn module of OpenFAST at some point, but we are not currently funded to do that. Until then, there are two workarounds to capture the effect:

  1. Model the tower torsion DOF through the nacelle-yaw DOF, i.e., by setting appropriate values for the nacelle inertia and yaw spring that mimic the effect of tower torsion.
  2. Model the entire tower using the SubDyn module rather using ElastoDyn (essentially locating the platform reference point in ElastoDyn at the yaw bearing). SubDyn captures not only bending, but also extension, shear, and torsion effects for circular members. The big limitation of SubDyn is that it is entirely linear, not including geometric nonlinearities that ElastoDyn considers for the tower (and blades). Modeling the entire tower / support structure in SubDyn is discussed in the SubDyn User’s Guide and Theory Manual.
    I hope that helps.

Best regards,