I’m trying to model Hywind Scotland spar turbine, here’s the problem:

I generated the Potential flow solution input data using Nemoh a code from Nantes polytechnic that does similar thing as WAMIT.
The results of these calculations are attached below (I’ll have to split to several posts)
In these graphs I am plotting my input files against the files for OC3 spar 5MW NREL irregular waves Spar.1 Spar.3 and Spar.hst files(dashed lines).
the platforms are different in sizes. So I thought that It’ll be fine. However my results have strong oscillations for most Platform dispacements.
OC3 spar is very deep 120m and diameter is 9.4
My spar for Hywind is only 77.6 m of draft but 14.4 meters in diameter and heavier about 10.7 thousand tons (7.5 for OC3)

Now, if I run 5MW NREL turbine for the same case as below, the agreement is amazing. Similar if I run my model with Spar hydrodynamic input files.
If I use my files the oscillations are just wrong. I’m attaching an example plot of my model with two sets of hydrodynamic input files, the ones I generated and the ones for OC3 model. The mean value is not the problem, the amplitude of motions is.

Below I will attach my Hydrodynamic coefficients (Xi for Beta=0), files attached below.

I was wondering… I have no access to WAMIT. Could somebody who does, please run one analysis for me to check my Nemoh results? Please! I would supply the mesh, and I only need one Beta, and 100 frequencies just to fit my calculations from Nemoh. I’m running late with my project for the uni…

I’m not sure I understand the problem you are describing. What are the various results you are plotting (I don’t understand the labels: ORE Catapult and OpenFAST, N-S, E-W, etc.)? What is the difference between HWS.png and SurgeSawySparInput.png?

I see some basic differences in your frequency domain solution of wave-excitation, wave radiation, and added mass, but nothing that stands out as concerning that would cause strong oscillatory response.

Could the problem be that the natural frequencies of your platform coalesce with the first-order wave-excitation frequency, unlike in OC3-Hywind spar, whereby the natural frequencies in surge/sway and pitch/roll are placed outside the frequency range of the first-order wave excitation?

This is very intersting! I’d better make sure these frequencies are in order!
I’m going to have a go this weekend.

In fact the SurgeSawySparInput.png is a plot of the results when I set PotFile to link to Spar files that you produced for OC3 Hywind turbine. And HWS is when I use the data I generated with Nemoh. Both sets are posted above in the .zip file. Nothing else was modified and the turbine is still the same model with my blades, and producing 6MW of power!
ORE Cataput is a UK based organization who share operational data from Hywind Scotland timeseries which I’m trying to match to with my model.
What I find amazing is that just by changing the hydrodynamic input files one gets completely different resoponses. It’s important because the pitch controller is very sensitive to this.
Is the process of placing the natural frequencies dealt with by setting the controll gains, like here (page 22) nrel.gov/docs/fy10osti/47535.pdf
or is it rather a more general turbine-support structure configuration issue?
Thank you again for your comments!

The natural frequencies of the platform modes will be dictated by the full system mass / inertia (including added mass) and stiffness (including hydrostatic and mooring), and will thus be based on the system configuration (geometry, material properties). If the platform-pitch or platform-surge natural frequencies are heavily impacted by the change in spar properties, then as you said, you may also have to modify the controller gains to ensure overall stable operation in above-rated operating conditions.

Thank you for your comments! It’s true, the natural frequencies from OpenFAST are nothing like the ones I believed to have modeled for!
To obtain the natural frequencies I ran a simulation with no waves or inflow and linearized with only the 6 DOFS of the platform enabled. Then I ran the matlab scripts fx_mbc3.m and campbell_diagram_data.m
Was this the right way to do this?

Now that I have to remodel the whole turbine, I would really like to have a quick way to view the mass matrix of the system in OpenFAST. is there a way to output it somehow?

The OpenFAST linearizations do not show the mass matrix directly, but this matrix can often be inferred from other matrices generated through the linearization process. A similar question was asked and answered in the following forum topic: http://forums.nrel.gov/t/openfast-2nd-order-linearization/2249/2.

It looks like there are many warnings about large angles, but it also said OpenFAST terminated normally. So, what do you mean that “other modes are not generated”? What does your ED_HWS_CASE_0_MODSHPSMBC.viz file look like? You can probably eliminate the warnings by reducing VTKLinScale in this mode-shape visualization file.

Thank you for your help.
I am attaching the files.
Now that I have read your previous reply though… Could it be that the other mode shapes are just zero? And that I am only seeing the yaw due to the ED Yaw moment I have in the inputs?
Maybe my linearization didn’t produce the results needed to visualize surge, sway, and other modes of the platform displacements?
I can see how I can find the

-M^-1K,-M^-1C

matrices, but in order to find

M^-1

do have to have loads acting on the platform listed as inputs?

It looks like you are only seeing one mode because VTKLinModes = 1 in your mode-shape visualization file.

Yes, you need to have loads acting on the platform as inputs to derive the 6x6 rigid-body mass matrix. These are available by setting LinInputs = 2 in the OpenFAST primary (*.fst) file and regenerating the linearization output.

Thank you for your suggestions!
I successfully generated VTK output for all modes of platform displacements but 1 and 2, surge and sway, but I’m sure that with some more trying it’ll be possible to get them as well.

I also obtained the M matrix from the linearization output. Thank you for your advice.
I used the subset of
D
transmission matrix produced by fx_mbc3.m from
ED Platform XYX forces/moments, node 1
into
ED Platform XYZ translation and rotation accelerations, node 1
I noticed the damping present in the model and stiffnesses affect the result.
At the moment I set the PotMod to 0 (and set the simple hydrodynamic coefficients to 0 as well) and I have good agreement with the mass matrix supplied by Equinor in terms of the mass of the total system. This matrix from Equinor is one without the part from mooring stiffness influence. the mooring stiffness matrix is supplied separately and I used it for the matrix of additional stiffness in Hydrodyn and disabled CompMooring. There is information that mooring creates 270 tons force downward, so I set this as preload in Hydrodyn in the -z direction.

I tried to balance with the mass of the platform to have as good a steady state as possible but there was an oscillation, so I set additional damping to 100 on all the diagonals but it didn’t help much. I had 0.4 m amplitude of the oscillation in surge when linearizing after 2000 seconds.

Taking into account how the mass matrix from OpenFAST was calculated from Newtons Law, I wanted to ask you:
1.Could I do anything else to get a more exact mass matrix? Rignt now I have differences of about 4000 kg between elements M(1,1), M(2,2) and 1000 kg between M(2,2) andM(3,3). Considering the mass is 11.5 thousand tons the error is not huge, but maybe I did something wrong?
Would constraing the DOFs and calculating one Platform DOF at a time change anything?

Should the specified preload be considered as mass added on top of the system mass and added to the element M(3,3)? this would make the results much worse…

Was my way of obtaining the mass matrix correct?

Is this right that to have as similar system as possible using the Force- acceleration transmission from linearization I would have to have some information about the damping in the real turbine which was not supplied. Is there a way to use the PotMode and somehow filter out the mass matrix of the model?
All I have is mass matrix, mooring stiffness matrix and eigenperiods.
My plan now is to adjust the inertias and mass distribution to match to the eigenperiods and mass matrix supplied using the mooring restoring matrix and linearizing to get the mass matrix of the model. but I’m afraid that later when I use PotMod=1 again for simulations the system will be wrong again.
What do you think I should do?

Regarding oscillations even after 2000 s, this is probably because you’ve zeroed the simple hydrodynamic coefficients (drag coefficients) and the remaining damping is low. You could always look at how the platform displacements are converging and use those (mean of the response) as initial conditions for a subsequent simulation.

Here are my answers to your questions:

Are you already generating your linearization output file with fine precision, e.g., OutFmt = “ES17.9E3”? If not, that may help improve the accuracy of the method.

No, the preload comes from the mooring system, but my understanding is that you want the rigid-body mass matrix of the FOWT without moorings.

Yes, that sounds correct.

I’m not sure I understand your question, but the force-acceleration transmission method should be independent of the damping specified.

thank you!
You’re right.
So it’s only the added hydrodynamic mass that is changing the mass matrix right? Because of how you designed Hydrodyn?
I think the errors in my mass matrix from linearization was due to the poor quality of my operating point selection. I added some more damping and the differences between Mi,i for the masses are gone.
Thank you, now I feel like I know what I’m doing.

I’m not sure I understand your question about the hydrodynamic added mass, but if I understand correctly, you are extracting the 6x6 rigid-body mass matrix of only the structural model because you are deriving this mass matrix by post-processing the ElastoDyn linearization output only (*.ED.lin). The hydrodynamic added mass is included in the full-system mass matrix when HydroDyn is enabled, but these terms are not visible in the ElastoDyn linearization output.

To clarify, I was using the rigid body mass matrix from ElastoDyn, yes. The masses I obtained on the diagonal or the first 3 rows and columns of the mass matrix correspond to my overall mass sum, so I think this went well.

I have to ask you about a detail of the Elastodyn input file.
Thanks to your kind remarks, I identified a lot of inconsistencies in my model and now I am rearranging the mass distribution of the whole system. I am getting close. I’ve reviewed most of the masses and made a sketch of the most important of the structural elements in order to get a better approximation of the values of the inertias. For the nacelle, I am considering the elements from the picture below: