I have been trying to model the OC3 - Phase I load case 5.3 in order to replicate the overturning moment spectra from the report ''OC3 – Benchmark Exercise of Aero-Elastic Offshore Wind Turbine Codes of 2007), however my results are always wrong by some order of magnitude.
The shape of the spectra does not seem totally wrong (see attached file where I add a t-domain simulation and the respective spectra), however it is slightly offset in the yyaxis (shape is similar, at least in average line, but the magnitude of the frequency components is offset in yy).
To compute the results i used a FFT algorithm (double-sided to define the spectra). My ultimate goal is then to replicate and validate the flexible foundation in the Phase II study.
I tried different things, such as increase of the Pierson-Moskowitz frequencies, different simulation times. But the problem is always the magnitude of the S_f components.
I also realized that this offset occurs when I divide the t-series output from FAST in order to convert to kNm.
Is it possible to confirm that these time domain overturning moment results are representative of what would be expected for the overturning moment (calculated at -20m mudline) in the 5.3 case?
I’ve attached a plot of the FAST time-series that was used to generate the PSD shown in Figure 13b of the paper you referred to (only a subset of the full time series is shown). Overall, this time series matches quite well to yours, with similar min, mean, max and std. Are you sure you are calculating the PSD correctly?
Many thanks for your answer. It helped a lot to separate the problem. The error was indeed in the interpretation of the Fourier transform. I was able now to get a PSD very similar to the one of OC3.
As I told you, my goal was to implement a AF (apparent fixity) foundation for further calculations. Regarding the Phase II, where the 5MW turbine is modeled using an AF foundation, the length of the AF beams was selected in order to replicate the soil conditions? (what I mean is, the foundation was tuned using the length of the beam in order to induce the same stiffness as the soil by using a tubular section similar to the monopile, D=6m and thickness=0.06m)
I have been trying to compute the PSD for these conditions (and also trying different stiffness values). The low frequency region approximates (up to 1Hz) approximates very well, but then I cannot ‘‘catch’’ the high frequency peak around 1.4Hz (Figure 7 for load case 4.2 with AF foundation - Paper Offshore Code Comparison Collaboration within IEA Wind Annex XXIII).
Do you know the reason for that high frequency oscillation? Is it because of the softer soil near the mudline? Is there any way to not miss those oscillation in my t-domain loading series?
I tried to decrease the sampling rate, and change the wave spectra frequencies, but still not able to compute that frequency peak.
Many thanks and regards
I’m glad you solved your original problem.
The AF model in OC3 Phase II was derived using a pile length, diameter, and thickness below the mudline that were tuned to mimic the overall response of the monopile above the mudline.
The frequency around 1.45 Hz in that Figure corresponds to the 2nd fore-aft natural frequency of the support structure (tower + pile), but with a rigid rotor-nacelle assembly. (See also Figure 4.) What frequency are you predicting the 2nd fore-aft natural frequency of the support structure to have?
Many thanks for your message. My expected second natural frequency was of 1.48 (through Bmodes).
I take the Stiff and Mass matrices to Bmodes from SubDyn and use it these in the calculations, using only the discretized properties of the tower (saw it in a post from before here). There I get some idea of the expected natural frequencies.
To model the t-series of loading I deactivated all the DOF of the RNA, leaving only True the ones from the tower and platform. Air density was put to ~0 (alternatively tried to turn off aerodynamic calculations).
This how an inverted pendulum should be modeled, correct? Is there any further consideration that I should take into the .dat files?
The spectra that I got for a flexible foundation with frequencies of 0.24 and 1.47 is below. There is a tiny peak near 1.5, but nothing close to the clear peak in Figure 7 of phase II report. I cannot even see this oscillating component in the respective time series.
Later, as i could not get the desired peak, I went back to the original (fixed foundation) monopile, were I get a 2nd FA frequency of 1.86 in Bmodes, but again when I try to model an inverted pendulum this frequency does not show in the spectra.
In the OC3 - phase I and II plots, the high frequency component looks clear when simulating regular waves, there is always this small oscillation of the loading, e.g Figure 5 of OC3 phase II or Figure 12 in OC3 phase I, around the “bigger periods”. This is associated to this frequency correct?
When I model loads, also with regular or irregular wave conditions, the low frequency component is there but not the high frequency one (I am sampling at 0.2s). See image below.
Do you think this may be happening because I am not modelling inverted pendulum well and rotor dynamics are meddling with the substructure frequency?
Also, I have not corrected yet the mode shape polynomial function in the time domain (as the difference was relatively low to the original mode shapes, from my little knowledge I am not expecting the problem to be in these as displacements are small), do you think this can be the reason?
Yes, an inverted pendulum can be modeled by disabling all but the tower and platform DOFs in ElastoDyn. I would disable aerodynamics (by setting CompInflow = CompAero = 0) rather than by setting the air density to zero.
From your PSD, it looks like you are picking up the correct frequency at around 1.5 Hz, but the excitation looks small or the damping is high. Do you have Craig-Bampton modes enabled in SubDyn? Have you tried varying the amount of structural damping to see its influence?
My apologies for the late reply.
It was strange to notice that when I change the critical damping conditions, there is no change to the spectra or the change is too small to notice (even at 100% crit. damping).
I have been trying to identify the cause, but was not able to. This is an unexpected response behaviour, correct?
However I noticed that I had a damping component in the hydrodyn file, and when i change it, the spectra does change significantly. (Imagine it is in the original hydrodyn file for the monopile, as i do not recall modifying it).
Considering the very small displacements, with a fixed foundation, should I expect this hydrodynamic component to have such large influence?
I’m not sure I can answer your questions without more information:
Which structural damping term are you changing? (There is both structural damping of the Craig-Bampton modes in SubDyn and structural damping of the tower-bending modes in ElastoDyn.)
What damping term are you changing in HydroDyn?
I was changing the term related to the AddBLin. It changes the dynamic results, however I could not confirm the DOF associated in the Hydrodyn manual. From the hydrodyn manual this matrix appears in the equations of motion of for the substructure.
There is also this information:
“While likely most useful for floating systems, these matrices can also be used for fixed-bottom systems; in both cases, the resulting load is applied at the WRP, which when HydroDyn is coupled to FAST, get applied to the platform in ElastoDyn (bypassing SubDyn for fixed-bottom systems).” (section 4.3.8 aand 6.4)
It appears on the third diagonal term for the [B] matrix (I would expect the influence of the damping on the overturning moment in this case to be on the B_11 or B_55 terms, but I am not sure of the convention used).
DO you think that there is any possibility that the Hydrodyn is by-passing the Subdyn using this matrix as a damping term ?
The sensitivity to the critical damping that I performed was on the Subdyn. I will check the Elastodyn also.
The AddBLin damping matrix in HydroDyn is used for fixed-bottom systems when HydroDyn and SubDyn are enabled. The (3,3) term adds damping in the heave direction. Section 5.4 of the draft SubDyn User’s Guide and Theory Manual (wind.nrel.gov/nwtc/docs/SubDyn_Manual.pdf) provides guidance on how to set AddBLin(3,3).
You could assess the contribution of other damping terms too (of the Craig-Bampton damping in SubDyn, of the tower-bending modes in ElastoDyn, and of the platform in HydroDyn (AddBLin)).
Many thanks for your message. Now I understand the heave damping component in Hydrodyn.
I continued to not be able to replicate the large peak for the inverted pendulum from phase II case 4.2. However, when I ran the fully couple model of case 5.2, I can obtain the exact same loading spectra. This makes me believe that for sure I am not replicating the inverted pendulum as it is supposed to. I needed to deactivate the DOFs, that I did, but do I need to change any other consideration regarding mass or inertia?
Many thanks for your support.
All you should need to do model the inverted pendulum case is to disable aerodynamics (CompAero = 0), disable the structural degrees of freedom in the rotor-nacelle assembly in ElastoDyn (FlapDOF1 = FlapDOF2 = EdgeDOF1 = DrTrDOF = GenDOF = YawDOF = False) and set the rotor speed to zero (RotSpeed = 0).
Many thanks for your message. That was what I was doing. Nevertheless, I have been checking and comparing my foundation with other, and the results for fully coupled operation seem more or less in accordance with other AF monopiles.
I just have (I hope) a final question regarding this topic. I updated the mode shape polynomial functions for the monopile and realized that I needed to decrease the delta_t to 0.001 (to avoid the small angle assumption violated error, which occurs right in <1s of simulation). This is bit of a constrain, as the model needs more time to be evaluated and I need to run a big number of evaluations for the analysis I intend to do.
Do you have any recommendation regarding any strategy, other than decreasing delta_t, to avoid the angle assumption error? I have been trying to play a bit with the blade angles and Elastodyn initial conditions for a couple of hours but did not have much progress.
Regards and thanks,
One should always use a time step that results in a “converged” solution. Did the time-domain solution change a lot when dropping the time step from whatever you had before to 0.001 s?
There are several topics/posts on this forum related with the small angle assumption violation warning and how to choose a proper time step in FAST. I suggest searching for what has been discussed in the past.
I came to realize that the TTdspFA at t=0 had a relevant influence in the capability to avoid the large displacement error right at the beginning of the simulation for the flexible foundation (maybe because I was always trying very demanding environmental operational states). I was able to improve convergence using a larger time step and avoid errors at t~0 by modifying the initial displacements.
I compared my mode shapes with the ones from the report , Modal Dynamics of Large Wind Turbines with Different Support Structures, and got similar mode shapes with slightly larger relative displacement at the tip in mode 1 and at mid-tower in mode 2. Particularly the mode 2 seems “very flexible” (figure red dots).
I read in other posts here that these polynomial functions shouldn’t be expected to influence the results in a significant way.
In fact I compared the results for the non-AF mode shapes and AF mode shapes using the same loading conditions and these did not show a significant influence in the results (even displacements). I Imagine because displacements are small, and as such small differences in relation to the tip are expected. Is this correct? Or I should expect relevant differences when modelling with the “corrected mode shapes” for the AF. (I am still trying to cover the theory documents, but I believe it will take me some time to grasp all of it…If I am ever able to)
In the attached SubDyn file FAST I am asking for outputs of “M4N1MKxe, M4N1MKye”, however the solution for these is always given as INVALID (see .out file). These are nodes below the mudline.
In SubDyn manual it refers that Hydrodyn (wtrdepth) is the one that sets the mudline for calculation of React. This means that no loads are calculated below it? I am not particularly interested in these loads, It is just to confirm that I am not making any mistake.
I’m not sure I understand what the source of the data is that you are plotting in your figures. Regardless, the tower mode shapes are important in the sense that if some linear combination of the mode shapes cannot represent the correct deflection of the beam for a given mode, than the solution may be effected if that mode is excited.
In SubDyn, output member 4 (M4) is not the same as MemberID 4. There is a mapping between output members and members IDs because SubDyn only allows 9 output members, but allows for unlimited members IDs. I see your SubDyn input file that you’ve only defined 3 output members (corresponding to MemberIDs 1, 3, and 4), so, output member 4 is invalid because it is not defined. See Appendix C of the draft SubDyn User’s Guide and Theory manual for more information: wind.nrel.gov/nwtc/docs/SubDyn_Manual.pdf.
SubDyn does allow you to output loads and motions of structural nodes that are below the mudline (for an apparent fixity model), but there are no hydrodynamic loads applied to members below the mudline.
These were the first two Fore-aft modes, original in the tower .dat file and after adding flexible foundation (blue and red respectively). My fault for not adding a proper explanation. It was in case you detected something that was “out-of-place” in the results. I was bit worried about the second mode (on the right, red markers) shape, but after checking other reports and dynamic results these seem within the expected shapes.
Thanks for the explanation on both the output members and loading schemes for the Subdyn output, I understand it now.
Best regards and again many thanks for all the help,
The red curve for the 2nd fore-aft tower mode looks a little odd in that I would expect the mode to have zero slope at the tower base (i.e. the tower-base is cantilevered to the monopile; the tower-base motion comes from the platform DOFs, not the tower DOFs). Perhaps the slope is zero, and it just doesn’t look that way in the image. Did you derive the tower modes for the flexible foundation using BModes and fitting a polynomial using the ModeShapePolyFitting.xls spreadsheet?
Many thanks, there was a mistake in the slope indeed. Yes I used the xls file and somehow I was able to mess the slope cell, so it was not being considered in an adequate way before. I corrected and the impact in the results (in terms of stability) is astounishing.
Everything looks smooth now.
A million thanks.
I tried to do a fatigue assesment and got surprisingly low stress amplitudes. To verify my model, I wanted to compare it with the results from load case 5.3 of the OC3 phase I. Unfortunately I do not have a turbulent wind field with the Mann model, so I generated a wind field with the Kaimal model instead. I created a turbulent wind field with TubSim v2.0, URef=18, IECturbc=14.8. For the hydrodynmaic loads, I used the linear wave theory, PM spectrum, Hs=6 and T=10. No wave-wind misalignment, no yaw error. I used the 5MW_OC3Mnpl_DLL_WTurb_WavesIrr from the r-test repository and only changed the HydroDyn input as mentioned above, changed the turbulent wind field in InflowWind, and added a line in ElastoDyn to output the tilting moment. I used openFAST v2.3.0 and the corresponding r-test input files (v2.3.0)
The max and min values of the tilting moment are much smaller than in the time series Jason previously shared. Could this be due solely to the different wind model? I did not expect such a difference in the variation of the overturning moment. Or is there something else that I might have missed?