FAST: Tower Model

Hello all,
Once again, thank you for your continuous support through this forum, its more than amazing how much it helps both sides improve their knowledge.

To try and make my post as short as possible, I will summarize what I want in the following lines:
I’m trying to create a 2D model for an Onshore Wind Energy Conversion Device Tower.

What is my model supposed to do?
Given the elasticity and section characteristics as well as forces and moments acting on the top of the tower, IT SHALL calculate the acceleration of the tower top.

What is my procedure?
I want to validate my model by the following steps:

  1. I take the Onshore 5MW turbine (provided within wind.nrel.gov/public/jjonkman/NR … Bsline5MW/ )
  2. I run FAST, obtain information related to forces and accelerations at Yaw Bearing.
  3. I apply the obtained forces from FAST onto my model, and let my model calculate the acceleration
  4. I compare the acceleration obtained by my model to the acceleration obtained by fast: if the match: perfect. if Not: debug and try again.
    PS: I set the structural damping in the tower file for all modes as 0.

What is my problem:
My model fails to get something similar to Fast’s acceleration.

Question:
Is the approach of considering the tower as a cantilever beam with a lumped mass on top (for Hub-nacelle assembly) correct? I used a beam element stiffness matrix with 3 DOFs per node, namely: Axial deformation, Transverse deformation and Rotation.
For the Mass matrix I utilized a consistent mass matrix.
Of course, I divided the tower into (N) beams and assembled the Global stiffness matrix, as the tower has a tappered section. It is worthy to say that the number of divisions performed on the tower is the same as those provided in the FAST .fst file

For the solution of the equations of motions: I formulated the well-known dynamic differential equation that relates Mass, acceleration, Damping, velocity, stiffness and displacement to the forces:
Ma + Cv + Kx = F(t)
for the special case of 0 damping, one gets:
Ma + Kx = F(t)

to solve for the acceleration, I used the Newmark Constant Acceleration Scheme.

I’m absolutely sure that there is no mistake in the stiffness, mass matrices as well as the solution scheme, as I tested it on a generic cantilever problem with F(t) and it yielded exact results.

Am I missing something? Any help is highly appreciated.

Best Regards and Uppermost Respects,
MJ

Dear Mahmoud,

I can foresee a couple problems with your approach:

  • Where is the rotor/nacelle-assembly (RNA) mass and inertia accounted for? The tower-top loads output by FAST intrinsicly include the inertial loads from the mass and inertia of the RNA. So, if the RNA mass and inertia is included in your isolated tower model, you’d be double booking these terms when you import the FAST tower-top loads. If the RNA mass and inertia is not included in your isolated tower model, than there is a potential for numerical convergence problems in your coupling approach; this is because the RNA inertial loads are acceleration dependent, making the coupling implicit (i.e., “F” depends on “a”). Moreover, if you included the mass of the RNA, but not the rotational inertia of the RNA in your isolated tower (you mentioned a “point mass”), this would also be a problem because the rotational inertia of the RNA for the NREL 5-MW turbine is large and will have a sizeable effect on the tower response.
  • It looks like your isolated tower model is linear whereas the tower model in FAST includes geometric nonlinearities, such as axial “shortening”, Coriolis and gyroscopic acceleration, etc. Differences in the tower model could result in different tower-top accelerations.

Best regards,

Dear Jason,

Thank you very much for your reply. I can see what you are pointing out, thus I will revise FAST’s source code again to get a better understanding of the underlying theories behind the analysis of a wind turbine.
It seems that taking out a piece of the whole turbine system and analyzing it won’t be that accurate.

Best Regards and Uppermost Respects,
MJ

Dear Jason

I am trying to analyze the structural response of an OWT with soil interaction using a FE package. The soil effects have been accounted for, in FAST8, using an equivalent fixity approach.
I have some doubts on the modelling of an equivalent support structure for the NREL 5MW OWT. Please do take a look at them:

  1. So far, i had modeled the OWT RNA, in my FE software (DNV-USFOS) as a point mass (rotor mass + nacelle mass = 3.5E5 kg) at the top of the tower. Does the nacelle mass (2.4E5 kg) include the hub mass, or should the hub mass (56,780 kg) be added separately, so that the total mass at the tower top becomes 4.1E5 kg?

  2. Also, from an earlier discussion in this post, i realized that, for a correct model, i have to include a rotary inertia corresponding to the RNA. How do i do this?

  3. I am planning to use an approach followed by ABS (2011), “Design Standards for Offshore Wind Farms” for transfer of equivalent quasi-static tower-top load from FAST to my FE model. ABS (2011) had taken the resultant maxima (Fx^2+Fy^2)^(1/2) of the tower base shear forces and moments, from an equivalent monopile in FAST for defining the load cases in their FE program. They had defined it as a reasonable approximation of the actual loading scenario.

    My idea is to use the same approach at the tower-top, wherein i would be defining the resultant of the horizontal shears (YawBrFxp, YawBrFyp) and moments (YawBrMxp, YawBrMyp) at the tower top
    as quasi-static loads for further analysis in FEM. Is this a reasonable approach? YawBrFzp and YawBrMzp at tower-top are ignored in the FE model, but will the RNA mass and rotary inertia compensate for this?

Sincerely

Abhinav

Dear Abhinav,

The hub mass of the NREL 5-MW turbine is included in the rotor mass of 110000 kg, so, the total RNA mass is 350000 kg.

I’m not sure that I understand your question about the rotational inertia of the RNA. Are you asking how to calculate the rotational inertia or how to use it in your FE model?

Regarding your third question, I’m not sure I can give more general guidance than I’ve already stated above, as I have not built such a model myself. I’ve seen publications in the literature that have addressed this topic, and I suggest you review those.

Best regards,

Dear Jason

Thank you for the clarification regarding the RNA mass. It has resolved my confusion.

My second question seems to be answered in a couple of previous posts of yours, http://forums.nrel.gov/t/tower-fore-aft-modes-shapes/193/1. sorry for not looking it up earlier. How can i use it in a FE model? What inertia values and references axes would i be using?

Please allow me to simplify my third question:
As FAST accounts for the mass and inertial effects of the RNA, if i were to use 6-component time series loads (from FAST) at the tower-top, i must eliminate the point mass and inertia from the FE model, otherwise, the loads would be double-booked. I hope my understanding is correct. However, if i were to use only the equivalent (resultant) or time-averaged static values of FA shear YawBrFxp and FA bending moment YawBrMyp at the tower-top, would I still have to account for the RNA mass, as i am ignoring the vertical forces from the FAST output?

Sincerely

Abhinav

Dear Abhinav,

I can’t really answer your second question because I don’t know anything about your FE model. My guess is you’d want to express the rotational inertia of the RNA about the top node of your FE tower model and specify this inertia at that node in the FE model.

Yes, you would double book the mass/inertia terms of the RNA if you considered them in your FE model while also including the full contribution (including inertia terms) of the tower-top loads from FAST. However such a coupling is also likely to cause numerical problems as I indicated in my Aug 06, 2014 post above. If you specified only the time-averaged load from FAST in your FE model, I would think you’d only be able to simulate the mean response in your FE model. I would think you’d want to specify the full 6 loads (3 forces, 3 moments) from FAST in your FE model; the axial (vertical) force is likely not too much different from the weight of the RNA (it will vary a bit due to inertia and aerodynamic loads), but you’d also neglect the yaw moment.

I hope that helps.

Best regards,

Dear Jason

Thank you for the information regarding the various load-transfer schemes and their related issues.

Sincerely

Abhinav

Dear Jason,

I was still wondering about modeling a wind turbine. I want to generate a FEM model (using simple bernoulli beam element, with their stiffness matrices)

Last time, the problem was that an operational wind turbine incorporates multiple effects that cannot be captured/modeled by a simple cantilever beam model. So I did the following:

  1. Modeled the NREL 5MW baseline onshore wind turbine as a parked wind turbine (blades at 90 degrees, and all instructions applied exactly as you explained in one of the forum posts)
  2. Obtained the tower top accelerations from FAST under a wind field generated by TurbSim.
  3. Obtained the moments and thrusts applied at tower top from fast (non rotating “actually it doesnt matter since no yaw is performed by the turbine”) from FAST
    ---- NOW on matlab ----
  4. Prepared the stiffness matrix of a cantilever beam (tappered, divided into N beams by using N+1 nodes)
    ---- and here comes the point ----
  5. Added variables to the assembled stiffness and mass and damping matrix to represent the effect of the RNA.
    The question is, whether the approach is ok or if I’m missing something. Furthermore, is it possible to “extract” the equivalent stiffness of the RNA at the tower top from FAST? if yes, how to do so?

Thank you very much.

Dear Mahmoud,

You haven’t supplied enough details on your approach for me to judge its correctness. I don’t see any problems at a high level, but the devil is always in the details. I’m assuming you are trying to apply the tower-top loads (forces and moments) from FAST to your FEM model? Of course, the comments from my posts above (particularly, my Aug 06, 2014 post) still apply.

With the FAST v7 linearization feature, you can derive the equivalent mass, damping, and stiffness matrices of the linearized wind turbine about a given operating point – see the FAST User’s Guide for more information. Of course, the linearized model contains the effects of all enabled degrees of freedom. Alternatively, you could probably derive the equivalent stiffness of the tower through a series of time-domain simulations e.g., by deriving the relationship between steady-state deflection and steady-state thrust.

Best regards,

Dear Jason,

I am kind of new to FAST and I am trying to reproduce the result for tubular tower in following report:http://www.nrel.gov/docs/fy05osti/36777.pdf
I studied the FAST manual and I know how to model the events and I did the same for 50 yr reoccurance for the 100m steel tubular tower.
Moreover, I change the tower properties in the tower related files.
The problem is, when I compare the results from FAST with the report the axial force and shear force (same direction with the wind) is kind of the same as the report but the moment is much different.
Is there any reason or is this giving you any clue to help me?

Best Regards

Dear Reza,

I’m not familiar with the report your referenced. Because you are considering the 50-yr event, presumably you are interested in direct wind loading on the tower? This feature is available in FAST v8 and most straightforwardly using AeroDyn v15; is that what you are using? What tower-drag modeling options did you enable?

Best regards,

Dear Jason,

First of all thanks for your quick reply.
check if I set the tower flags correctly:
WakeMod: 1
TwrPotent: 1
TwrShadow: False
TwrAero: True
AIDrag: False
TIDrag: False (when I turn them on the ElastoDyn cannot converge)

However I added the tower points in the AeroDyn file and now have the answer.

Best Regards

Dear Reza,

For the 50-yr condition, I would expect the rotor to be parked or idling. In this situation, you can also disable the wake calculation altogether by setting WakeMod = 0. The other settings are then fine.

Best regards,

Thank you Jason
It helped a lot

Dear Jason

I simulate the parked wind turbine with EWM 50 yr reoccurance and constant wind speed of 42m/s.
In the constant wind speed, the Fy, Mx, and Mz are oscilating. I would like to understand it fully. Can you help me with a short explanation or give me any reference to study?

Best Regards

Dear Reza,

I don’t know enough about what you’re doing to answer your question. What Mx are you plotting (blade-root bending, rotor torque, tower-base bending)? What structural DOFs have you enabled?

Best regards,

Dear Jason
I am really really sorry, I was in my world.
I am simulating the 100m on 5MW shore wind turbine. The wind speed is constant and it is 59m/s.
The plotted Mx is the bending moment at the tower base. When I use EWM the Mx doesn’t oscillate.
I am simulating the parked wind turbine with the generator off, pitch angle of 90 and 0 rpm and all other conditions mentioned in the FAST manual.
once again sorry for previous non sense question .
Best Regards

Dear Reza,

Is there a yaw error? I’m aware of a potential blade-edgewise instability that can occur when the NREL 5-MW turbine is parked with all blades feathered to 90 degrees in high winds with yaw errors between 20-40 degrees. You can read about this in Section 6.2.1 of my PhD thesis-turned NREL report: onlinelibrary.wiley.com/doi/10.1002/we.442/pdf.

If the shaft axis is nominally aligned with the wind direction, I’m not sure what is causing the oscillations. I would suggest that you simplify the model a bit (e.g. by disabling structural DOFs) to help identify what features of the model are driving the oscillations.

Best regards,

Dear Jason,

I would like to ask if it is able for FAST to calculate the wind load on the turbine tower?

If it is, could you tell me which subroutine is capable of doing this?

Thank you very much in advance.

Best regards,

Yingyi Liu