I am using the 5MW_TLP_DLL_WTurb_WavesIrr_WavesMulti file for simulation in OpenFAST.
I am trying to comparison of motion RAO (Figure 26-RAO WAMIT 27m) calculated by Denis Matha in “Model Development and Loads Analyis of an Offshore Wind Turbine on a Tension Leg Platform, with a Comparison to other Floating Concepts”.
The way I do it is calculating motion RAO using white noise spectrum, with reference to the paper("Investigation of Response Amplitude Operators for Floating Offshore Wind Turbines) and forum posts
[Method]
I used 8000 seconds out of a 10000second simulation.
wave time series → Sxx (spectral density ) using Blackman-Tukey
ouput response(surge, pitch) → Sxy (spectral density) using Blackman-Tukey
motion RAO = Sxy/Sxx
For surge motion RAO, I think the peak values match and y(RAO)-values for each frequency appear to be influenced by smoothing.
However, For pitch motion RAO, the magnitude of the peak values varies significantly.
Is there something I am forgetting or doing wrong?
Please give suggestions on how to analyze.
It looks like your natural frequencies are lining up well between your results and Matha’s results for both the surge and pitch modes, but the level of damping is quite different, with your results showing significantly less damping. This was done many years ago now, and I don’t recall the details, but I see in section 3.5 of Matha’s thesis-turned NREL report that it says his results include aerodynamics enabled with the rotor spinning at a constant 12.1 rpm at 11 m/s. I don’t see that your results include aerodynamics, but the lack of aerodynamic damping in your model could explain the difference.
First of all, I appreciate your compassionate response.
Modified content: Aerodynamic effects were considered when the rotor rotates at a constant speed of 12.1 rpm at 11 m/s.
Result: Aerodynamic damping effects have been confirmed. It was effective in Pitch RAO, but the peak value(y) decreased in Surge RAO.
I was comparing the pitch WAMIT 27m in Figure 26 with WJ(OpenFAST).
I think there could be a difference between the value transformed from time domain to frequency domain and the WAMIT(spoke 27m).
I am reviewing whether there are any errors in value settings and trying various approaches.
First of all, I resolved the error based on your advice. Thank you.
I have a question regarding the values in Denis Matha’s “Model Development and Load Analysis of Offshore Wind Turbines Installed on Tension Support Platforms, and Comparison with Other Floating Concepts.”
First, looking at the previous results, the peak frequency of the pitch motion RAO matches at 1.8 [rad/s] (Figure 26).
However, the natural frequency in Table 6 is 0.2211[Hz], which corresponds to 1.3892[rad/s].
When the full system(platform+turbine) is considered as rigid and a static offset test (free decay) is performed, the results match the natural frequencies in Table 6.
The values I calculated match the natural frequency values in Table 6, but they differ from the RAO(5) Pitch WAMIT 27m in Figure 26.
The natural frequency of RAO(5) Pitch WAMIT 27m in Figure 26 should also be the same, as it is calculated as a rigid body.
Are the analysis conditions for Table 6 and wamit (Figure 26) different?
I will attach the values I calculated along with the relevant data (Table 6 & Figure 26).
This work was done many years ago by Denis Matha and I don’t recall the details now. But my understanding is that the pure WAMIT analysis assumes a fully rigid system whereas the FAST model has structural flexibilities, like the tower, which shifts the platform-pitch frequency lower. The FAST 27m result in Figure 26–with a peak around 1.4 rad/s–seems quite consistent with the pitch natural frequency in Table 6 of 1.39 rad/s.
I attached the spokes to the NREL TLP.
Spoke Lengths 18 [m]
Spoke Width 2 [m]
Spoke height 2 [m]
I performed hydrodynamic calculations considering the spokes using ANSYS AQWA (C.O.G is 0,0,0).
I converted the ouput format to WAMIT format and used it in OpenFAST.
Based on the principle dimensions provided in Matha’s paper, i calculated the freeboard and set the tower attachment point at 9 [m].
Q1.
As a result of the free decay test, I think the motion for surge and heave is normal, but pitch is unusual.
It started from an initial displacement of 0.02 [deg], but the motion occurs up to -0.025 [deg].
Q2.
The results of the free decay test using OpenFAST converge to a non-zero value. (in pitch motion graph)
So, I am trying to determine whether this is due to an error in the settings related to rotational motion.
Q3.
If RNA is assumed to br rigid and the Tower’s flexibility is considered, the free decay test result(pitch) show multiple peaks caused by tower’s flexibility.
Do these peaks each have a physical meaning?
Q4.
I am studying how to apply the Morison equation while reading the HydroDyn manual.
HydroDyn includes a Ca value, but is it not applied to avoid duplication with the hydrodynamic value(from WAMIT)?
I have attached the relevant materials and the results.
It appears from your pitch free-decay motion that there is an offset of -0.02 m at equilibrium (with a rigid tower), likely from the overhanging weight of the RNA. This is why the magnitude of the lowest negative value is larger than the magnitude of the highest positive value.
See 1.
These are likely the tower-bending modes coupling with the pitch motion and showing up in the pitch free-decay.
When PropPot = TRUE, only the viscous terms from the strip-theory solution are enabled for that member, so, while the drag coefficient (Cd) is used, the values of the added mass (Ca), pressure (Cp), and buoyancy (Cb) coefficients are not used for that member.
First of all, Thank you for letting me know.
I have gained an understanding of the free decay results and the application of morison’s equation in OpenFAST.
I have a question before peforming calculations using QTF data(sum frequency) in NREL TLP.
[Current Status]
I used the NREL TLP base file. (no spokes attached & no freeboard)
For the mooring analysis, I applied the MAP module and the MoorDyn module, which uses the lumped mass method.
The results of the analysis using the MAP module differed from those using MoorDyn.
Environmental conditions: Hs 4[m], Peak period 8.4[s] and no wind
[Result]
Pre-tension values are different.
When the MAP module is applied using the mooring analysis method, the response is generally larger than that of the MoorDyn module. (Slack phenomenon abserved.)
Q1.
Since both modules will start from the hydrostatic equilibrium state, I’m wondering why there is a difference in the responses.
After resolving this issue, I plan to proceed with a model that considers spokes and freeboard.
I also plan to use OrcaFlex for validation.
Therefore, I would like to use MoorDyn, which uses the same lumped mass method.
Q2. ★★
The set-down phenomenon caused by surge motion, which is a charateristic of TLP , is no observed.
I think the reason is that wind forced were not considered.
Q3. ★★
I think that, due to the strong axial stiffness of TLP, the heave displacement in the + direction relative to the hydrostatic equilibrium state should be small.
However, the graph shows that the heave displacements in both the + and - directions relative to the hydrostatic equilibrium state are similar.
I would appreciate your expert opinion.
I have attached the relevant materials and the results.
You mention that MAP and MoorDyn have different pretensions, but the tension results in your plots are quite similar; can you clarify?
You mention the presence of slack lines in one model and not another; can you clarify how you are seeing that?
In a TLP, heave motion can be the result of set-down (due to large surge motion) and elastic stretching of the lines. In your results, I would guess the latter is dominating because the surge motion is not large.
I do see a difference in the heave range between your MAP and MoorDyn results; I’m curious if this is related to the line damping you’ve included in MoorDyn. How did you determine BA, and if you reduce BA, do the results from MAP and MoorDyn get closer?
First of all, I appreciate your compassionate response.
There was an mistake in the analysis of the phenomenon.
the answer is as follows.
The pre-tension for both analysis methods is 3,833 [kN], which is the same value.
I thought both models exnibited slack.
This is because, I thought that when the tension approached zero in the dynamic analysis, it was slack.
Is it not allowed to perform the analysis for the NREL TLP model at environmental condition(Hs 4[m], Tp 8.4[s])?
I am looking for the minimum allowable tension for a quantitative analysis.
Thanks to you, my questions have been answered.
Looking at the data from the transient range(until 500s), I found that the surge had a displacement of 5[m], corresponding to a set-down in the heave.
Based on other studies, the heave displacement in the + direction relative to the hydrostatic equilibrium position of the TLP was smaller than that in the - direction.
However, the environmental conditions were different for me.
Even when i changed Jonswap’s peak period, the surge motion was not significant.
(ex. 17s, 14s, 11s, 8.4s)
l’m wondering which physical parameter or condition dominates the surge motion.
I analyzed the heave motion while reducing the BA value from its default value of 6.0e6.
Environmental conditions: Hs 4[m], Peak period 8.4[s] and no wind.
Reducing the BA value makes the heave motion similar to that obtained using MAP.
Is it okay to change the BA(axial damping) value included in the provided NREL TLP file?
There is no mention of this in Matha’s paper.
Thanks for clarifying. Here are my responses to your points:
I agree that a line tension of zero is indicative of a slack-line event, but I would also expect the transition from slack back to taut to also cause a lot of ringing, which don’t necessarily see in your response.
First-order wave excitation will result in surge oscillations about zero mean. To increase the mean would require second-order wave excitation or wind loading.
I’m not actually sure where the BA value originated in the MIT/NREL TLP model, but see that the value of 6.0e6 N-s is the value included in the r-test. In the work of Matha et al, the original quasi-static mooring model of FAST v6 was used (the predecessor to MAP++), not MoorDyn (MAP++ and MoorDyn did not exist at that time).
[BA]
I learned that the MAP++ and MoorDyn modules did not exist in the past.
My goal is to perform an analysis by attaching the SPOKE while considering its hydrodynamics and the Tower’s free board.
I plan to perform verification using OrcaFlex to validate the model. Therefore, I performed a free decay test to verify both software programs.
The verification of RNA, Tower, hydrodynamic coefficients, and hydrostatic has been completed.
I have also excluded the influence of Morison.
Therefore, I believe that the BA is the cause of the difference in the free decay test results.
In the case of OpenFAST, I’m really wondering why there’s so much heave damping.
Regarding (2), which value of BA are you using and is the mooring damping you are using in MoorDyn consistent with the damping you are using in OrcaFlex?
First, I am using a BA value of 6.0e6 (the default value).
Additionally, in OrcaFlex, it is not possible to directly input the axial damping (N.s) of the mooring system.
The input for OrcaFlex is as follows.
I tried various methods to reconcile the results of the Free decay test obtained using the two software programs. (change of BA 6.0e6, 3.0e6, -1, etc)
But the results didn’t change.
If you remove the Morison term in OpenFAST, isn’t it the case that only radiation damping is considered? ★★
[heave motion]
In the case of OrcaFlex, I am considering only radiation damping at this test. However, it seems there is additional damping in the case of OpenFAST.
[pitch motion]
In the case of OpenFAST, the decay period is different. I think this is a different system.
I have checked the factors that affect the pitch natural frequency, but I’m still not sure. (RNA mass information, Tower and hydrodynamic coefficients)
In the forum, I found that the mass and diameter values in the mooring inputs for OpenFAST and OrcaFlex are different. However, I don’t think this is the cause of the issue.
From your OrcaFlex screenshot, it looks like OrcaFlex supports Rayleigh damping, but that you’ve disabled it. This would equivalent to using BA/-zeta = 0.0 in MoorDyn. OrcaFlex may also have numerical damping from its time integration, but I’m not familiar with the details.
For your OpenFAST model, it looks like you’ve included radiation damping in HydroDyn and tendon drag and structural damping in MoorDyn. You also have the tower degrees of freedom enabled, which may have modal structural damping (but you have not shared your tower file to confirm).
I’m not sure why the OpenFAST solution is diverging, but I would first check that the numerics are converged by trying smaller time steps (DT) and/or adding a correction step (NumCrctn > 0) in the OpenFAST primary (.fst) input file.
I’m curious if the small differences in the coupled free-decay periods are the result of differences in the tower models between OpenFAST and OrcaFlex. To isolate this, you could try disabling the tower degrees of freedom (DOFs) in both OpenFAST and OrcaFlex. Furthermore, you could isolate the free-decay to specific platform DOFs.
First of all, I have some good news.
The delay in my response was due to organizing the data.
I had only considered the line’s Rayleigh damping, but your response helped me understand the structural damping of the tower.
I followed the steps below in order.
[Possible causes of the Free decay result problem]
Pretension → 4,133 [kN] for OpenFAST and 4,110 [kN] for OrcaFlex
Hydrodynamic coefficients → Verified
Hydrostatic restoring → Verified
Internal damping of the line (Tower ★ & Tendon)
[Step 1. Free decay test for the full rigid system]
Removed the tower flexibility.
Heave decay result did not matched.
→ Heave decay problem is tower’s internal damping
Pitch decay result is matched.
→ Pitch decay problem was tower’s internal damping
Based on these results, I thought I should first match the line’s internal damping between the two software and then consider the tower’s flexibility.
[Step 3. Matching the values of OpenFAST’s tower modal structural damping with OrcaFlex’s Rayleigh damping]
OpenFAST represents the tower using the coefficients of a 6th-order polynomial. OrcaFlex models using line elements (lumed mass model).
Therefore, OrcaFlex cannot apply a damping ratio to each tower mode.
Due to my limited understanding of ElastoDyn, I used only the natural frequency of the 1st tower fore-aft mode to determine the Rayleigh damping in OrcaFlex. the value of TwrFADmp(1) is used.
The natural frequencies of the tower fore-aft (FA) and side-side(SS) were taken from Matha’s paper.
As a result, the pitch decay response and the natural period matched.
I assumed that the 1st tower F-A mode would have the most significant influence on pitch motion. Is this a reasonable approach?
I have a question regarding the transient response of surge motion in OpenFAST.
I tried to resolve the issue by looking through many posts on the forum.
The content is as follows.
[In TLP Surge motion]
To determine the appropriate start time for using data, I compared the surge motion results from OpenFAST with those from OrcaFlex, where a transient period can be explicitly specified.
As shown in the attached figure, the surge motion in OpenFAST appears to approach equilibrium only after approximately 4,000 s.
Based on the input settings and the characteristics of a TLP, I believe that the slow decay of surge motion may be related to low damping in surge.
However, I’m not certain whether this long transient period is physically reasonable or caused by modeling or parameter settings.
I also tested the simulation using pre-calculated hydrostatic equilibrium initial conditions, but the results remained the same.
Could you please share your opinion on whether such a long transient period is expected for a TLP system in OpenFAST, or if there are specific parameters I should check?
Thank you very much for your time.
