the Creative Commons Attribution 4.0 License.
the Creative Commons Attribution 4.0 License.
| 30 Jun 2025
On the development of a hardware-in-the-loop wind tunnel setup to study the aerodynamic response of floating offshore wind turbines
Abstract. In floating wind turbines, the met-ocean conditions lead to motions of the floater affecting the rotor aerodynamic loads, which in return influence the motion of the floater, in a highly coupled way. Numerical design tools have proven to fail to predict some aerodynamic phenomena, such as the increase in thrust variation caused by unsteady effects. Thus, experimental testing is essential for tuning and validating these codes. Hybrid testing in wind tunnels, by reproducing numerically and actuating the floater motions while measuring aerodynamic loads on a physical scale turbine model, overcomes the scaling issues of traditional wave basin tests allowing a higher fidelity in the reproduction of the aerodynamics. This work presents the development of a hybrid hardware-in-the-loop setup designed to study the aerodynamic response of floating wind turbines in wind tunnels. A scale model of a multi-megawatt floating wind turbine is mounted on top of a six degrees-of-freedom hexapod robot. The full coupling of aerodynamic and floater dynamics is obtained with a hardware-in-the-loop approach with force-feedback-motion-actuation architecture. The rotor loads measured on the physical rotor are fed into a floater dynamic numerical simulator which calculates the motion in real-time and actuates it through a moving platform called hexapod. Key outcomes include the development of a hardware-in-the-loop numerical model with a force correction method to cope with scaling effects and an assessment procedure to verify the simulator, correction model, and measurement-actuation chain. The aerodynamic effects on the motion response are preliminarily investigated on a 10 MW floating concept, with direct estimation of the rotor aerodynamic damping showing a 210 % increase of damping in pitch with the turbine in operation. The capability of testing combined wind and wave cases is also demonstrated, setting the framework for future studies.
- Preprint
(4160 KB) - Metadata XML
- BibTeX
- EndNote
Status: open (until 28 Jul 2025)
-
RC1: 'Comment on wes-2025-100', Anonymous Referee #1, 11 Jul 2025
reply
Dear Authors,
I was invited by the Associate Editor to review your article and was pleased to accept, as the topic is of significant interest to me. I dedicated a considerable amount of time to its review. As you state, there are currently very few setups capable of conducting hybrid wind tunnel experiments for floating wind turbines and the development of such systems could be important for advancing floating wind technology. Therefore, I believe the topic can be of interest to the research community and the readers of this journal.
However, the manuscript requires substantial revisions to be technically sound, reproducible, and impactful.
As it stands, the article primarily describes the development of a HIL setup as a preparatory step for your future experiments, with limited contribution to the understanding of coupled dynamics in floating wind turbines. While I fully understand the need to establish a foundation for future work, a publication must provide clear value and insights for the reader.
General comments
I recommend that you clearly state, in the introduction, the relevance of your work, its main objectives, and its novelty compared to previous studies. In other words, the readers should understand what they can gain from the article. To support this, I suggest expanding the literature review so that gaps in the current state of the art become more evident, and your contribution can be better contextualized.
A second general point concerns the technical accuracy in some parts of the manuscript. I believe responding to the Specific Comments can help improve this aspect, but I also encourage greater care in ensuring accuracy and clarity in the article.
Specific comments
- Abstract: “feedback control”. Which kind of control? Generator speed/power?
- Abstract: “Numerical design tools have proven to fail… by unsteady effects”. Relatively recent research and community efforts have shown that numerical models are quite effective in capturing the thrust response of floating wind turbines especially without active rotor speed and blade pitch control as in the present work. I ask the authors to be clearer on what the numerical codes fail to predict or remove this kind of statements if they are not true or not meaningful for the present work.
- Abstract: rephrase the sentence “by reproducing … scale turbine model” because it’s not clear.
- Abstract: “higher fidelity” with respect to what?
- Abstract: “with direct estimation … with the turbine in operation”. Is this damping increment found also in simulations? Was it expected?
- L28: after “due to the interaction of the wind turbine blades with their own wake” you should cite work on this topic. Examples are [1, 2] but feel free to cite others.
- L31-34: “the results … numerical models”. How likely it is that the unsteadiness is due to the wind turbine aerodynamics and not something related to the experimental setup (like bending of blades, tower, load cell flexibility and related resonances)?
- L37-38: “This was done … a wind turbine”. Did they use experimental results to investigate aerodynamics modeling? If not, maybe the citations are not relevant here.
- L48: “A Reynolds’ similarity law…” Is Reynolds scaling (i.e., keeping the Reynolds constant) even possible?
- L62: “for specific studies”. Can you be more specific about the kind of studies where prescribed motions are useful?
- L72: “complex interactions” interactions of what?
- L79-81: “Hybrid wind tunnel testing … wind tunnel”. I'd say that these references are not very meaningful here. There many studies of wind turbine aerodynamics / wake using prescribed motions, and the two cited by the authors are some of them. However, the rest of the discussion is on HIL experiments, and I think it would be better to focus on those.
- L87: “correcting the measured force”. At this point of the article, it is not clear what force correction is. I would say something like "estimating rotor loads from the available measurements".
- L109: Move “The relevant properties are reported in Table 1” before the sentence “A complete …”.
- Table 1: Make length scale and velocity scale consistent with what is reported in the text body.
- Table 1: replace “Tilt” with “Shaft tilt”.
- Figure 2: in the caption it is said “The experimental points are averaged over two sets.” Two sets of what? Please explain it in the text. You should briefly explain how the characterization was performed. I get that thrust is the main driver of the floating wind turbine response and it was the main objective of the scaling but also torque is important because the turbine is in the end a power generating machine. Can you also show the torque corresponding to the measured thrust reported in Fig. 2?
- Figure 2: the figure displays only thrust. Can you show the corresponding values of torque?
- L115: change “in the previous study (Taruffi et al., 2024b).” with “in the previous study of Taruffi et al., 2024b.”
- L117: “without tracking errors” How this relates to amplitude of motion? At least you have a limit on the amplitude imposed by the maximum acceleration of 1g.
- L118: “most load conditions” Is it true? A wave period of 8s, which is not so uncommon corresponds to a mode-scale frequency of 6.3Hz which is above the range where the hexapod works ok. The maximum model scale acceleration of 9 m/s2 corresponds to a full-scale acceleration of 0.54 m/s2 (0.06g) which seems quite less than what a floating wind turbine can experience. Can you be more accurate on what you mean with "most load conditions"? Maybe it is more accurate to say something like "load conditions with mild waves" etc.
- L124: “rotor force correction process”. Again, I think it’s not clear here what you mean with correction.
- L141: “NREL 5MW … to the scale”. Add reference documents for the two wind turbine designs and be more accurate on what you mean with adjusting the operating parameters (rated wind speed, working TSR) and why this is acceptable (e.g. the optimal TSR of the IEA 15MW is different from the one of the DTU 10MW, thus if you run your rotor at the TSR of the IEA 15MW the local angle of attack is suboptimal for the rotor design).
- Table 3: Again, be consistent with the text and other tables when reporting the scale factors. Moreover, the first two lines of Table 3 are repeated from Table 1. I suggest removing them from Table 1.
- Section 3: Which is the delay in the measurement-actuation chain? Where it mostly comes from? (e.g. the hexapod motion controller). Which is the transfer function between motion setpoint and actual motion of the hexapod?
- L174-178: “Differently from Belloli … are not measured”. The same is done in [3] which is the development of Belloli 2020, for the same reason mentioned here. It is also said “We preferred to have the load cell at tower-top rather than at tower base (like in Belloli et al. 2020) because in this way the aerodynamic components are larger fractions of the signal, and to avoid introducing deformability at tower base that would reduce the frequency of the turbine first fore-aft mode.”
- L194-197: “In this work … in real time”. I think the content of these sentences is evident and doesn't need any further explanation.
- L210-211: “the angular accelerations … in the current setup”. Can you estimate the error on Aerodynamic loads you are committing by neglecting these terms? Or alternatively how much Tin+Tgrav change by neglecting the angular accelerations?
- L213: “delays between sensors signals and hexapod feedback”. This sentence is vague, and you should explain here that there is a delay (which is quantified afterwards) and explain which are the consequences of this delay.
- L215-216: “de-synchronization of signals ... acceleration inputs”. I think the authors must explain more clearly why they didn't use filters. I suspect the first bending frequencies of the scale model (tower and blade FW) could be at around 10 Hz. The forces resulting from the vibrations of the scale model components are measured by the load and are transferred to the numerical model of the platform causing it to respond at the same frequency of the flexible vibrations. This normally gives rise to a loop that could make the HIL system unstable. Why it's not the case? Is it related to the motion tracking performance of the hexapod which maybe filters out and does not track motions at high frequency. I ask the author to provide the transfer function (in a plot or also by means of values at some frequencies reported in the text) of the hexapod setpoint-->actual movement in terms of magnitude and phase.
- L218: “This reduces errors … using a different sensor”. The meaning of this sentence is not clear, rephrase it.
- L259: “Prior to wind tunnel … the results”. It seems a reasonable step and I don't see the need to give this advice to other people. You can simply say that The HIL setup was verified in cases without wind.
- L268: “it corresponds to testing in Open-Loop”. It doesn’t. It corresponds only if the robot tracks perfectly the setpoint from the numerical model. However, the robot tracking performance is not proved anywhere. I suggest removing this part of the sentence.
- L269: “FAST”. Can you be more accurate and specify which version of (Open)FAST you are using?
- L276: “stiffness”. I get that you must adjust damping since HydroDyn uses a more complex model, but why you had to adjust stiffness? how was the mooring stiffness matrix obtained? How did you conduct the tuning? Are final values close to initial guesses or not? Is the difference expected?
- Figure 4: “at full-scale”. In the results, report everything at full scale or at model scale.
- L288: “neglecting the non-rotating rotor drag”. I get what you mean, and I think it's better to say that the aerodynamic force developed by the non-rotating rotor is small.
- L294: “with the only exception of the yaw case”. The yaw response isn't that bad.
- L296-298: “This, as explained … negligible effects”. Rephrase, it's not clear. I think it's better to spend more line of text here to clearly explain why force correction on yaw is more critical than for the other degrees of freedom.
- L298-299: “Moreover, this study … has a limited impact”. I'd say that the study focuses on cases with aligned wind and waves, where the yaw response is expected to be small, thus the accuracy for this DOF can be lower than the others, especially surge, pitch and heave.
- L300: “the results obtained in the wind tests are accurate”. This is not true. In these tests, the aerodynamic force resulting after correction is ideally zero. If it is introduced in the numerical model with a delay, it doesn't produce any error in the system response because it has zero magnitude. In these tests it was only possible to assess the accuracy of the force correction procedure but from its results it's not possible to say that tests with wind are accurate.
- Figure 6: I think this figure doesn't add much compared to figure 5 and can be removed.
- Section 3.3.3: I think this section should be used to show the FRF motion setpoint-->actual motion
- L302: “within a limit”. Can you quantify this limit?
- L305: “sensibly higher for the force sensor”. How much higher?
- L330-331: rephrase the sentence “The thrust force … in the direction of thrust” because it’s not clear.
- L335: “where the part of the rotor … causing the damping effect”. This explanation is difficult to understand especially when it says “the opposite happens simultaneously on the other side, causing the damping effect.”. Please rephrase it and make it more robust and less speculative.
- L350: “this is also the region where the effects on wind are visible”. The figure only shows the response with wind but not the case of waves without wind (like in Fig. 7-8) so it's impossible to tell at which frequencies the aerodynamic loads have a meaningful effect.
- L355-356: “The dynamic response to wind and waves … this kind of study”. This sentence is very broad. You must add reference of the literature that you mention and explain how your findings compare to theirs. Even better, you should add to figures which is the expected response (e.g., from numerical models).
- Figure 9: Can you do some averaging when computing spectra (e.g., with Pwelch), otherwise it's difficult to identify the features of the spectra.
- L359-360: “only one wind tunnel HIL setup exists”. Not true. These setups are scarce, but there are others, for example [4]. As the authors acknowledge, the published work on this kind of setups is not much, so I suggest them to extend their survey of literature that now is a bit limited.
- L368-370: “By measuring forces at the tower top… force ratio”. Not true, this is done in [3]. The novelty here is using measured accelerations.
- L373: “unsteady aerodynamic phenomena”. I invite you to caution about these phenomena you mention because it has not been clearly proven yet that they exist and how the HIL setup presented in this work may eventually contribute to their understanding.
- L374-377: “the versatility of … aerodynamic phenomena”. These sentences are broad and a bit useless because the flexibility of use mentioned by the authors is not demonstrated in the article. I suggest recalling the main results of the article instead.
- L378-379: “A preliminary study … degrees of freedom”. Be more accurate on what you found.
- L393: what you mean with “theoretically measured”?
- L405: where it has been “already calculated”?
Technical corrections
- Title: I suggest removing “On the”.
- Abstract: replace “met-ocean conditions” with “wind and wave excitation”.
- L11: remove “dynamic”.
- L12: “hexapod” it is said before that the turbine is mounted on a hexapod robot so there is no need to specify it again.
- L20: remove “deeper”
- L24: “FOWTs experience (rigid-body) motions”.
- L61: remove the comma after “rotor”.
- L106: “a velocity scale of 3”. It should be “1:3” to be consistent with the geometry scale reported at line 105.
- L106: replaced “scaled” with “designed”.
- L127: replace “coupling” with “coupler”.
- L131: replace “controller” with “computer” or “machine”.
- L131: replace “perform the hardware in the loop” with “run the hardware-in-the-loop” controller.
- L132-133: “(DAQ)” and “(HMI)” These acronyms are not used anywhere else in the text so I think it is unnecessary to define them.
- L134: “real-time machine”.
- L155: remove brackets and add “whose scaling is governed by” after “motion frequency”.
- L181: replace “. This” with “; this”.
- L182: add “between the physical and numerical parts of the experiment.” after “mismatch”.
- L183: rephrase the sentence “due to manufacturing … the mass scale” because it is hard to read.
- L187: replace “read” with “measured”.
- L187: “the measurements are heavily compromised”. I would say that is not possible to use the measurements as a feedback.
- L232: what does “derived” mean here?
- L234: what does “flexibly targeted” mean?
- L235: remove “The general equation of motion … hereafter” because it’s a repetition of the sentence afterwards.
- L238: I suggest removing “R2 is the quadratic damping (here not modelled)” and remove R2 from Eq. 4 to avoid confusion.
- L240: I suggest removing “Fmoor is the mooring load (here not modelled)” and remove Fmoor from Eq. 4 to avoid confusion.
- L242: remove “The viscous effects … damping matrices”.
- L252: remove “(SS_Fitting)”
- L253: replace “the radiation calculation” with “F_{hydro}”
- L262: replace “3” with “three”.
- L264: "loop" instead of using italics recall what is the loop.
- L273: remove “The wave response … in wave conditions”.
- L320: replace “in wind conditions” with “with wind blowing”.
- L323: “at rated conditions”. Can you recall the wind speed and rotor speed?
- L331: replace “no control is present” with “rotor speed and blade pitch are fixed”.
- L358: remove “(HIL)”.
- L359: remove “(FOWTs)”.
- L366: add “Because of the additional complexity introduced by closing the loop,” before “Investigating the effect…”.
- L367: remove “therefore”.
- L415: “3” is it “Eq. 3”?
- Appendix A2: I suggest replacing “rotating frame” with “moving frame”. In wind turbine, “rotating frame” is usually used to identify the coordinate system that rotates together with the rotor.
- L426: capital R in “Rotation” should be small.
- L433 and 434: replace “rotating” with “rotated”.
References
[1] Papi, F., Jonkman, J., Robertson, A., and Bianchini, A.: Going beyond BEM with BEM: an insight into dynamic inflow effects on floating wind turbines, Wind Energ. Sci., 9, 1069–1088, https://doi.org/10.5194/wes-9-1069-2024, 2024.
[2] Schulz, C. W., Netzband, S., Özinan, U., Cheng, P. W., and Abdel-Maksoud, M.: Wind turbine rotors in surge motion: new insights into unsteady aerodynamics of floating offshore wind turbines (FOWTs) from experiments and simulations, Wind Energ. Sci., 9, 665–695, https://doi.org/10.5194/wes-9-665-2024, 2024.
[3] Fontanella, A., Facchinetti, A., Daka, E., and Belloli, M.: Modeling the coupled aero-hydro-servo-dynamic response of 15 MW floating wind turbines with wind tunnel hardware in the loop. Renewable Energy, Volume 219, Part 1. https://doi.org/10.1016/j.renene.2023.119442, 2023.
[4] Jiang, Z., Wen, B., Chen, G., Tian, X., Li, J., Ouyang, D., Peng, Z., Dong, Y., and Zhou, G.: Real-time hybrid test method for floating wind turbines: Focusing on the aerodynamic load identification. Journal of Ocean Engineering and Science. https://doi.org/10.1016/j.joes.2024.06.002, 2024.
Citation: https://doi.org/10.5194/wes-2025-100-RC1 -
RC2: 'Comment on wes-2025-100', Anonymous Referee #2, 21 Jul 2025
reply
Dear Authors,
It was very interesting to review this paper. Empirical data plays a key role in increasing our understanding of the physics of such complex systems, and for the development of numerical tools.
The paper is well written and well structured, and there is no doubt that a considerable amount of work has been put in the development of a HIL system for testing of offshore wind turbines in a wind tunnel.
The introduction states that this work primarily focuses on the verification and validation of the setup. This is an important step in the development of the setup, but one would expect more results and a more thorough analysis of the results to be really convinced about the verification and validation of the setup.
Also, the introduction is interesting but lacks clarity regarding the advantages and the limitations of model testing in a wave basin compared to a wind tunnel. In the abstract , it is stated that the wind tunnel HIL tests overcome the scaling issues of traditional wave basin tests. I would think that this is false, and that the only advantage of wind tunnel tests compared to wave basin tests is in the quality of the wind. Also, additional challenges arise for the wind tunnel tests due to the non-froude scaling and the very limited mass of the RNA.
The paper in its currents state needs more work but will be worth publishing once the two above comments have been addressed. The specific comments below are meant to help you to improve the paper regarding my two comments above, but additional results and discussions are needed regarding the validation and verification.
Specific comments:
L38-40: One of the main advantages of laboratory testing compared to field testing is that the environment is controlled in a laboratory. I think that this should be stated as one of the reasons why laboratory testing is often preferred.
L51-55: It is important to give a clear review of basin tests with performance scaling.
- Why do you state that an arbitrary velocity scaling is used while my impression is that Froude scaling is often used. See for example Bredmose (2017). By using Froude scaling in basin tests, the problem about mismatch in the scaling of the gravity is for example avoided.
- How do you define low quality wind flow. The wind quality is the main difference between basin tests and wind tunnel tests and deserves therefore more than one sentence.
- The mass of the RNA is difficult to match, but is it really not achievable? Already in 2017, Bredmose was only 12% above the specification. But even with a RNA that is a bit heavier than desired, it is still possible to achieve the correct centre of gravity, which is correct for rigid models.
L95: The hybrid setup: Can you add more details about the quality of the flow in the wind tunnel (spatial and temporal variation, capabilities, …), since this is supposed to be better for wind tunnels compared to basins. And this is one of the main “selling” arguments for HIL wind tunnel tests.
L117: The hexapod: Since the main objective of the paper is to verify and validate the setup, one expects more information about the tracking errors of the hexapod. “… without tracking errors”: Does this mean 0 errors, or negligible? A bode plot with amplitude and phase would be very valuable here.
L143: I do not agree with the statement that the main driver for hybrid tests in wind tunnels is to overcome the Fn-Rn scale conflict. The same issue is present in HIL wind tunnel tests and therefore, performance scaling is also used in wind tunnels. The main driver for HIL wind tunnel tests is in the quality of the wind. It is also important to highlight the challenge that arises in HIL wind tunnel tests, due to the non-Froude scaling. While this is not a problem in basin since Fn scaling is used there in combination with performance scaling.
L189: Please explain how you arrive at the mismatch in ratio of 150. The mismatch in ratio between the aero and the inertial forces is the same as the Renolds mismatch (given in table 2) times the mass mismatch (10). The mismatch in ration between aero and gravitational is the same as the mismatch in Rn*Fn*mass mismatch (10).
L221: Floater Dynamics. Since the objective of the paper is verification and validation of the HIL setup, this section about the numerical model should be exhaustive. The information is scarce, and it is difficult to know what is exactly included in the numerical model.
- Equation 4: Added mass on the LHS is a part of the F_Hydro. Same for F_rad in L454.
- L246: Is it only radiation damping or should it be radiation loads?
- L247: Please clarify what is meant by wave diffraction since Faltinsen and Newman have different interpretation for this term.
- What about second order wave loads?
- L456: What is F_diff,2 and why is it not included in A7.
L266: Floater simulator: If the objective is to verify and validate, then one should compare FAST and the Simulink with the same parameters and verify for a good agreement. By tuning the parameters for agreement, you are hiding possible errors in the numerical model.
Figure5: The agreement in pitch is not so good while pitch is one of the key DOF for Floating wind turbines. Would it be possible to use the angular pitch acceleration derived from the accelerometers at tower base and top?
L309: What do you do with the estimated latency? Can you give additional explanations on why you believe that you can split the delay evenly between motion and communication. Other laboratories have tried to compensate the delay, see for example [2] Have you tried similar?
Technical Corrections:
L34: Can you add a reference to the statement about premature maintenance.
L49: Rn does not compromise the hydrodynamic loads (viscous effects are correctly scaled), but it compromises the gravity related hydrodynamic loads.
L51: Is it correct to state that an arbitrary velocity scaling factor is used in performance scaled tests? My impression is that in basin tests, performance scaling is used in combination with Fn scaling. See for example [1].
Figure2: Did you have to adjust the blade pitch angle to achieve the desired thrust?
L146: Even with non-Froude, inertial phenomena can be correctly reproduced. The problem is that the balance between inertial and gravitational loads is not correctly reproduced.
Table 3: Please add Froude mismatch
L184: The same challenge would arise also for custom made components. The challenge is mainly due to the complexity of RNA and the need for sensors and actuators.
References
[1] H. Bredmose et al., “The Triple Spar campaign: Model tests of a 10MW floating wind turbine with waves, wind and pitch control,” Energy Procedia, vol. 137, pp. 58–76, Oct. 2017, doi: 10.1016/j.egypro.2017.10.334.
[2] Zhihao Jiang, Binrong Wen, Gang Chen, Xinliang Tian, Jun Li, Danxue Ouyang, Zhike Peng, Yehong Dong, Guiyong Zhou, Real-time hybrid test method for floating wind turbines: Focusing on the aerodynamic load identification, Journal of Ocean Engineering and Science, 2024,
Citation: https://doi.org/10.5194/wes-2025-100-RC2
Viewed
| HTML | XML | Total | BibTeX | EndNote | |
|---|---|---|---|---|---|
| 142 | 34 | 21 | 197 | 7 | 10 |
- HTML: 142
- PDF: 34
- XML: 21
- Total: 197
- BibTeX: 7
- EndNote: 10
Viewed (geographical distribution)
| Country | # | Views | % |
|---|
| Total: | 0 |
| HTML: | 0 |
| PDF: | 0 |
| XML: | 0 |
- 1