Dear Community,
I am a new user in this forum and before posting I checked if the topic of soil-structure interaction for wind turbines was already discussed, but I couldn´t find any item of information.
I am a civil engineer and I investigate the influence of the type of soil on the dynamic behaviour of the wind turbine. Concerning the model I focus on the tower-foundation-soil system using a coupled finite-boundary element method, while everything atop the tower-top will be lumped in a point mass element. However in order to compute the time-histories of the aerodynamics loads acting on the top of the system I choose FAST, instead of too simplified formulas provided by standards or guidelines.
To cut a long story short, if my tower is extended with the foundation-soil system, the dynamic behaviour of the turbine is influenced (and that´s actually the heart of the investigation), therefore the mode shapes differ from the ones of a cantilevered beam. Neither the deflection nor the slope at the tower foot is zero, and I guess I need a complete 6th-order polynomial to approximate the functions, but the tower input assumes automatically that the 0th and1st term are zero.

Or maybe I just compute the mode shapes for an identical structure without the foundation-soil part (as a cantilevered beam) I run FAST and then apply the loads to the complete top mass-tower-foundation-soil system, though with a bit of incongruity.

As an overall idea, is it wrong to compute the loads with FAST and apply them to an external model, based on finite element method and boundary element method? A coupled wave propagation problem-rotor aerodynamics analysis is not meant to be a task of my research, but I would like to refine my approach. In general the aim of my transient analysis is to investigate the influence of the type of soil on the dynamic response of the turbine subjected to concurrent wind and earthquake loads. I have an appropriate preparation on seismic and structural issues (or at least that is what I studied for) but unfortunately not a good insight into aerodynamics.

I appreciate any help and I hope the topic falls in your interest.
Best regards,

Dipl. Ing. Francesca Taddei

RWTH-Aachen University
Lehrstuhl für Baustatik und Baudynamik
Tel: +49 (0)241 / 80 25089

I’m familiar with the question you raise. The basic issue is that you want to represent the correct mode shape based on the proper boundary conditions, but FAST requires that you input the mode shape of a cantilevered beam. Regardless of FAST’s requirement, it is possible to input the correct mode shape in FAST. Here’s how:

In the correct mode shape, the tower base has an associated translation and rotation, unlike a cantilevered beam, which has no tower-base motion. In FAST, however, the tower-base motion (including tower-base translation and rotation) is represented through the 6 DOFs of the platform. Also, the tower DOFs are defined relative to this motion, so, in effect are cantilevered to the platform motion. If you place a line tangent to the correct mode shape at the tower-base, and project the mode shape onto this tangent line, the mode shape will appear cantilevered to the line. It is this “pseudo-cantilevered” mode shape that you want to specify in FAST. We’ve developed the MS Excel workbook ModeShapePolyFitting.xls, included with FAST, to do this projection and calculate the mode shapes needed by FAST. Applying this procedure, we’ve been able to use FAST to properly model towers with different tower-base boundary conditions, including idealized free-free beams, “coupled springs” and “distributed springs” representations of the foundation, and floating platforms.

If there is little direct coupling between the soil and the turbine (which is typical), then it should be OK to decouple the analysis–i.e., take the tower-base loads from FAST and apply them to the FE soil code. The key thing to keep in mind, however, is that the natural frequencies of the tower as modeled in FAST should be tuned so as to match the proper natural frequencies of the real system (including the influence of the soil). For this, it often OK to include the soil effect in FAST with a simplified model, such as the “apparent fixity”, “coupled springs”, or “distributed springs” representations of the foundation.

Dear Jason,
thank you very much!It really helps. I did not check the file ModeShapePolyFitting.xls before, but now I got the idea.
Just a short question: if in the Fast input file at the section PLATFORM I enter PtfmModel=1 (onshore), am I able to introduce a set of stiffness-damping coefficients (as a model for the foundation-soil system) for the 6 DOFs of the platform?
Sorry for asking maybe obvious questions but I did not find an example for the file containing the platform properties.

Thank you in advance.
Best regards,

Dipl. Ing. Francesca Taddei

RWTH-Aachen University
Lehrstuhl für Baustatik und Baudynamik
Tel: +49 (0)241 / 80 25089

Yes, you can model the stiffness and damping of the foundation in FAST when PtfmModel = 1 (onshore). This is what is known as the “coupled springs” representation of the foundation. Because FAST has up to 6 DOFs at the tower-base, the stiffness and damping can be represented by 6x6 stiffness and damping matrices for a linear representation (nonlinear stiffness and damping can also be implemented).

In the current version of FAST, however, you can only define this foundation stiffness and damping through the user-defined routine UserPtfmLd, which requires a recompile of FAST to use. Once UserPtfmLd has been written and FAST recompiled, to have FAST use UserPtfmLd during a simulation, set PtfmLdMod to 1 in the FAST platform input file. Take a look at UserPtfmLd and you’ll see that the sample we provide already includes a placeholder for a linear 6x6 stiffness and damping (and added mass if you so desire) matrices.

that´s great. I´ll try this out and I´ll compare it with my finite elements/boundary elements model for the tower+foundation+soil system.
Thank you a lot.

Greetings,

Dipl. Ing. Francesca Taddei

RWTH-Aachen University
Lehrstuhl für Baustatik und Baudynamik
Tel: +49 (0)241 / 80 25089

It says in the ModeShapePolyFitting.xls “It is best to enter the known slop from a program such as BModes”, I couldn’t find the slope information in BModes output. Would you show me the way?
Thanks,
Sail

For every mode output by BModes, BModes supplies the transverse displacements (s-s disp and f-a disp), transverse slopes (s-s slope and f-a slope), and twist as a function of distance along the beam (span_loc). The “known slope” you want to enter into ModeShapePolyFitting.xls is the slope (s-s slope or f-a slope) at the bottom of the beam (span_loc = 0)

FYI: We now recommend that BModes be used in place of Modes. BModes employs a higher fidelity model than Modes and also includes the option for tower-top inertia (in addition to point mass) and tower-base DOFs for foundation flexibility or floating platforms. The biggest issue with BModes is it doesn’t directly output the polynomial coefficients needed by FAST. We’ve supplied the “ModeShapePolyFitting.xls” spreadsheet in the FAST archive to aid in this effort of deriving the polynomial coefficients.

Dear Jason,
I am modelling the 1.5-MW 3-bladed WT (described in the WindPACT Turbine Rotor Design Study) with my finite element-boundary element code.
I need to know the dimensions of the tapered tower.
Reading the report of the “WindPACT Turbine Rotor Design Study” I found the following geometry from the “Table 6.2-Summary of Properties of the Baseline and Task #5 Final Configurations” :

Thick_top= 0.0087 m
Thick_base= 0.0174 m
D_base= 5.663 m
D_top= 2.823 m
while from the Test13.fst file I found out that:
Flexible Heigth of the tower= H= 82.39 m
I assumed the following steel properties:
Ex= 2.00E+11 N/m^2
Ey= 2.00E+11 N/m^2
G= 7.70E+10 N/m^2
rho= 7.85E+03 kg/m^3
Poisson´s ratio= 3.00E-01
I assumed at first that for top and base diameters the author meant the internal diameters of the top and base annulus and I calculated the distributed properties of the tower (included in the attachment).
The volume of the tower is H*(A_top+A_base)/2=1.60E+01 m^3 while the tower mass is Volume*rho=1.25E+05 kg
My results are in disagreement with the ones recorder in Baseline_Tower.dat.
Even if I assume that the top and base diameters are meant as external diameters, the results do not match the example in The CertTest folder.
However I decide to substitute the values in the file Baseline_Tower.dat with my results and when I run the program I obtain a funky value of the tower mass:
(From the fsm file)
Tower Mass (kg) 116393.037
which is different than 1.25E+05 kg.
Am I doing something wrong? Maybe I confound myself but I do not see the mistake in my calculations.
Are my assumptions for the material properties right? Maybe the density of the steel is assumed different than rho=7.85E+03 kg/m^3.

Thank you for any help.
Best Regards,

Dipl. Ing. Francesca Taddei

RWTH-Aachen University
Lehrstuhl für Baustatik und Baudynamik
Tel: +49 (0)241 / 80 25089

I wasn’t the one who originally derived the distributed tower properties of the WindPACT 1.5-MW baseline wind turbine supplied with the FAST archive. Instead, these came from data supplied by GEC (one of the leads on the WindPACT studies). I quickly checked the source of the distributed tower properties and see that they were derived from slightly different values of the diameter/thickness than you report. Here are the values that were used:

base diam = 5.61 m
top diam = 2.82 m
base thickness = 0.01756 m
top thickness = 0.00882 m

The other values you report, including the flexible tower height and steel properties are the correct, except that the density was increased by 5% to account for paint, bolts, flanges, etc. that are not accounted for in the thickness data.

Table 6-2 in the “WindPACT Turbine Rotor Design Study” refers to configuration 1.5A08C01V03cAdm. The version used to derive the WindPACT 1.5-MW model supplied with the FAST archive was configuration 1.5A08V07adm, which is a newer version according to the configuration ID described in Table 2-4 of the same report. I’m not sure why the report lists properties of an older version.

I haven’t checked all of the equations in your attachment, but hopefully the changes in tower values described above will enable you to reproduce the distributed tower properties supplied with the FAST archive.

I am looking at putting monopiles or other fixed foundation under 13.2MW turbine in GoM shallow waters. You said FAST can do “coupled springs” or “distributed springs” representations of the foundation. I am not sure I got these functions of FAST so far. And how to use that?

Another question is, up to now HydroDyn can use Linear Wave Theory to model waves. For shallow waters, is there any mathematical model existing in HydroDyn such as Stream Function Theory for calculation.

It is not possible in the current version of FAST to directly enable a coupled springs or distributed springs model solely through settings in the input file. Instead, a slight customization (requiring recompile) of FAST is currently required. The coupled springs model can be implemented in FAST through the UserPtfmLd routine, which allows one to apply a discrete load at the platform reference point. The distributed springs model can be implemented in FAST through the UserTwrLd routine, which allows one to apply distributed loads (loads per unit length) along the flexible tower. Please find the UserPtfmLd and UserTwrLd routines that I created for modeling the coupled springs and distributed springs models in the IEA Wind Task 23 Subtask 2 Offshore Code Comparison Collaboration (OC3) project attached. You should be able to use/adapt these attachments to introduce soil flexibility into your FAST model.

The current version of HydroDyn for monopiles accounts for regular or irregular linear waves (with or without stretching) and sea currents and uses the relative form of Morison’s equation for the load calculation. Because the module does not currently have a higher-order wave kinematics model built into it (e.g., Stream Function theory), if you want to model severe regular waves, you would have to use GH Bladed or some other wave kinematics code to generate the higher-order wave kinematics data before running a simulation with HydroDyn. The WaveMod = 4 feature was added so that HydroDyn could read data generated externally by GH Bladed or an equivalent code.

I’m currently working with UserPtfmLd in UserSubs to implement mudline flexibility and damping in FAST for the offshore wind turbine (supported by a monopile). Can you explain what role the platform added mass (PtfmAM) plays in the mudline calculations, and what it defines in a fixed-bottom context? I found that without specifying a rather large platform added mass, my simulations won’t run - and depending on the mass, it adds a “second” first natural frequency to my structure (e.g. a second peak in a Fourier Transform of free vibration aside from the first natural frequency of ~0.3 Hz).

The platform added mass (PtfmAM) within UserPtfmLd() could be used to represent acceleration-dependent loads transfered to the pile from the soil. However, I’ve seen soil models that ignore these terms; so, you can probably set PtfmAM to zero.

Enabling the platform degrees-of-freedom (DOFs) will introduce additional natural modes in the FAST model. My guess is the model is not running because these new modes are at a high enough frequency that that the structural time step must be reduced to resolve them without numerical instability. (And introducing PtfmAM lowers the natural frequency of these modes, making the model numerical stable.) See my post dated Nov 4, 2010 in the forum topic found here for rules of thumb on how to choose proper time steps: http://forums.nrel.gov/t/error-in-fast-working-in-adams/298/1.

a short question: If I add soil matrices (stiffness and damping) in the user-subroutine, in the case of soft soil, FAST delivers very good and reasonable results. If I implement very stiff soil, which means very high stiffness coefficients, the program aborts saying…

I though about a convergence problem and I reduced the time step, succeeding with my stiff soil without problem.
Therefore my conclusion is that, if I add a very stiff soil and with it a high stiffness, the algorithm starts to have problem with convergence as the system is characterized somehow by high frequencies (and low periods) and the time step need to be reduced.
Does it make sense? Is that the physical reason?

If I want to incorporate p-y curve in the UserTwrLd subroutine, do I have to specify the resistance (p) equation related to displacement (y) just right at the TwrNode position where the displacement vector is calculated?

Did somebody have such examples to share with me? I don’t know whether can we do this in this routine.