the Creative Commons Attribution 4.0 License.
the Creative Commons Attribution 4.0 License.
Predicting the Onset of Dynamic Stall on Large Wind Turbines
Abstract. This study addresses the challenge of predicting dynamic stall on wind turbine airfoils, focusing on the development of a reducedorder model applicable to thick airfoils (t/c > 0.21). Utilizing a Delayed DetachedEddy simulation of a pitching FFAW3211 airfoil at Re = 15 M, our analysis identifies the transition from the primary instability phase to the vortex formation stage as a critical aspect of dynamic stall. By examining the dynamic time scales, we observed a tenfold increase in the growth rate of the shear layer height during the transition of these stages. The stall delays attributed to these stages are substantially dependent on the airfoil's camber distribution and the location of the maximum thickness. We discovered that the LeadingEdge SuctionParameter (LESP) proposed by Ramesh et al. (2014) for thin airfoils is also helpful in predicting the onset of the vortex formation stage for thick airfoils. Based on this finding, we propose a MidChord SuctionParameter (MCSP), that is more effective for wind turbine airfoils. The MCSP exhibits a breakdown in magnitude at the onset of the vortex formation stage and deep stall.
 Preprint
(1323 KB)  Metadata XML
 BibTeX
 EndNote
Status: closed

RC1: 'Comment on wes2023159', Anonymous Referee #1, 06 Dec 2023
The manuscript addresses the need for a reducedorder model to predict the onset of dynamic stall appearing on the wind turbine (WT) blades which inherently have thick airfoil profiles. The authors state that the effective way of delaying the dynamic stall is to develop new airfoil profiles that maximize the angle of attack initiating the stall vortex formation, which is possible by predicting the time scales of dynamic stall stages via reducedorder models. This constitutes the motivation of the study.
To monitor the dynamic stall stages on thick airfoils, the authors propose a MidChord SuctionParameter (MCSP), derived from the LeadingEdge SuctionParameter (LESP) which was previously introduced in the literature and already used for thin airfoils. In the paper, a pitching FFAW3211 airfoil is simulated by DDES, and both LESP and MCSP are tested using the acquired unsteady data. The paper concludes that both parameters are helpful for predicting the onset of dynamic stall stages whereas MCSP seems more reliable as the flow separation occurs at midchord on the thick airfoils.
The study addresses the relevant topics of the journal of WES, and the proposed method (MCSP) might be of widespread interest to the wind energy community and the WES readers. The manuscript, in general, is clear and wellorganized in terms of structure and presenting the results. The literature survey is quite sufficient. However, with the presented results and discussion, it is difficult to come to a conclusion that the MCSP is a more robust criterion for WT airfoils than LESP. Including some additional results that show flow visualizations investigating the flow separation regions and so as the dynamic stall would be beneficial to support the conclusions about the proposed method.
Please find some questions and suggestions to the manuscript as follows:
 Dynamic stall is difficult to capture even with robust CFD techniques. DDES is an accurate and effective tool for massively separated flows where large eddies dominate the field; however, its inherent defects are likely to appear in case of smallscale eddies generated inside the shear layers. DDES models all structures inside the boundary layer until the flow detachment. Thus, the use of DDES may adversely affect the prediction of primary instability and vortex formation stages. Instead, LES or modified versions of DDES (such as IDDES, ZDES, DDESSLA) would be more appropriate (and of course more costly) for this problem. Could you please give reasons for selecting the DDES method with relevant references?
 In Section 2.1, there is a sentence between Lines 105107 stating that the kw SST model provides turbulence closure for 3D RANS in the attached flow regions of DDES. “In the attached flow” phrase might be misleading here because indeed the same equations are solved everywhere in the domain, but an additional switching mechanism in the dissipation term of the kequation makes DDES behave differently in the regions apart from the attached flow. So, I suggest simply removing the phrase “in the attached flow” from the sentence.
 Please clearly define alpha_* and alpha_DS before specifying them, like for alpha_SS.
 In Section 3.2, it is underlined several times that for the investigated case flow is attached near the leading edge and dynamic stall vortex is formed at midchord where the flow detaches. Accordingly, MCSP is calculated using the region of 0.45<x/c<0.55. In addition, (as already mentioned above) MCSP is found by the authors more reliable than LESP due to the flow separation regions. A reader may need to see the flow behavior on the suction side, especially in the LESP and MCSP regions, through some instantaneous flow snapshots. So, it is highly suggested to include additional figures demonstrating the vortex formations and shear layer heights in this section.
 In Section 3.2, the paragraph between lines 232 and 238 discusses the cl results of DDES shown in Fig.4. However, there is no discussion about the BEM results. Please compare DDES and BEM results presented in Fig.4.
 On page 13, the stall delay attributed to the primary instability stage (Deltat_1^*) is compared with literature data using the A/k term of Equation 6. Then, the effect of airfoil geometry on this stall delay is analyzed with the help of Fig.6. In this discussion part, I think there is no need to mention f(beta) as it is not related to the airfoil geometry and also it is not even used in the comparisons. So, I simply suggest removing f(beta) from Equation 6 (also from the text) and directly writing Equation 6 as Deltat_1^*=A/k.
 In Line 277, it says that the MCSP intersects the LESP at t_SS^*. Is there any meaning of this intersection?
 Please find some suggestions/corrections related to grammar, formatting, and typos as follows:
 Line 2: The letter “s” in “simulation” should be capitalized.
 Line 8: No need to abbreviate “LeadingEdge SuctionParameter”.
 Lines 32: “Brunner et al., 2021” should be in parentheses.
 Line 34: “thicknesstochord ratio” instead of “thicknesstochordratio”
 Line 63: What is ROM? Is it BEM?
 Line 68: It should be only “LESP_crit” instead of “critical LESP_crit”.
 Lines 7173: The sentence starting with “LESP_crit is …” may be rewritten as “Since LESP_crit is a function of the airfoil geometry and Reynolds number, using thin airfoil theory it can be predicted with the first term of …”
 Line 77: “stated” instead of “state,”
 Line 90: “thicknesstochord” (or t/c directly) instead of “heighttochord”
 Line 99: “are” instead of “is”
 Line 112: No need to abbreviate “finite volume”.
 Line 137: “CFD” instead of “computational fluid dynamics (CFD)”
 Line 143: Is “alpha” here the induction factor? If so, it should be “a”.
 Line 158: “Model” instead of “model”
 Line 159: “BLM” instead of “BeddoesLeishman Model (BLM)”
 Line 160: “Leishman and Beddoes, 1989” should be in parentheses.
 Line 188: “nu” should be written as its symbol.
 In the caption of Table 4, it should be “ratio” instead of “ration”.
 Line 253: “in Fig. 6” instead of “in 6”
 Line 267: I think instead of “t_SS^*”, “t_DS^*” is intended to be given here, right? If so, it would be better to check “12.35”.
 Line 292: No comma before “demonstrates”.
 Line 296: “thick” is repeated.
 Line 390: The initials of the first author name are missing.
Citation: https://doi.org/10.5194/wes2023159RC1 
RC2: 'Comment on wes2023159', Anonymous Referee #2, 07 Dec 2023
general comments
The submitted preprint proposes the use of a new parameter, namely the MidChord SuctionParameter (MCSP), to better predict dynamic stall on a wind turbine blade when using reduced order models.
The paper starts with a for the most part wellwritten intro giving the reader a good overview of the problem at hand and the state of the art. The subject is also of significance to both industrial and scientific wind energy community and presents a new approach. The figures are well presented and rich in material without also exceeding the capability of the reader to extract information from them. The relevance of the paper is also well justfied in the intro and the formating is generaly very good.
However, the paper leaves the reader somewhat thirsty for more results, comparison and clarification. If the goal of the paper is to convince the community to switch to the MCSP approach, a more thorough study should be conducted. The reader also has the idea the paper was not enough reviewed before being submitted, as there are some inconsistencies, which I relay in the comments below.
The remainder of this review is structured in the "specific comments" and "technical corrections," which each list remarks that the authors should address by expanding or clarifying directly into their paper.
specific comments the names of the boundary conditions are inconsistent between table 1 and figure 1
 it is somewhat confusing that much of the theory presented in the introduction deals with thin airfoils (thin airfoil theory), perhaps because they emanate from the helicopter community, while the paper deals with thick airfoils. A few sentences on the implications of this, regardless of dynamic stall, should be given to the reader
 it remains unclear to the reader without OpenFOAM experience what the boundary conditions actually are: inletOutlet, cyclicAMI, "fixedValue 1e9", nutlowReWallFunction, etc. It is nice to give the OpenFOAMspeccific names for reproducibility, but it should also be explained to the reader what these are in a few words for each of the boundary conditions listed in the table.
 the same goes for Table 2
 although it is not a problem, why was symmetry instead of a periodic boundary condition used on the sides of the airfoil and what are the implications of this?
 the authors mention they developed a new cyclic boundary condition, but no info is given on what is new about it; is cyclic/cyclicAMI not a standard part of OpenFOAM? (although a citation is given, it should not be expected that the reader reads that other paper just to know this and a short sentence should thus clarify the novelty of their sliding interface)
 the mesh at both sides the sliding interface should be shown, especially if the detaching vortices cross it (do they?)
 the text mentions that a constant growth factor is applied for the entire domain, but behind the trailing edge there is at least one jump with approx 2.0 growth, which is not a problem per se but does contradict the text, can you comment on this? Does this come from the relaxation of the normal extrusion parameters (is it hyperbolic extrusion?) or was the mesh adapted by hand (perhaps to project it on the sliding interface?)?
 can you also give delta(z) as z^+ values?
 which software or algorithm was used to generate the mesh?
 what type of shielding model was used for the DDES; with what parameters?
 page 7, lines approx. 163179: this paragraph should be reformulated, it is quite hard to follow and understand what the authors mean
 also, to which coefficient do you refer when your write "normal coefficients"?
 line 188 "freestream velocity" is quite confusing after having two pages of theory which also covers helicopter rotor measurement; you should remind the reader that you are talking about your 2.5D pitching airfoil model. This becomes even more confusing around line 195 where you prevent selfintersection but nondimensionalizing the time with a parameter that depends on U_inf/c (which is a constant, right?). This also needs to be clarified.
 the authors should comment on the impact of having simulated only 3 cycles (of which only 2 are taken for the results) and of the averaging process. It is known that dynamic stall is an process that can vary strongly between cycles and the state of the art is to group the dynamic stall force cycles into multiple clusters, each exhibiting different separation characteristics.
 overall, the introduction and the long description of the BEM and it's (and other reduced order models) lack of proper handling of dynamic stall give the reader the impression that the goal of the LESP and MCSP is to improve such models, but then the LESP and MCSP are obtained from a DDES simulation, and not used for the BEM results that are presented (or if the LESP or MCSP are used to correct the BEM results of figure 4, it is not clearly stated). All this needs to be made clearer in the paper by explaining how the reduced order model (BEM here?) toolchain is modified and what is the difference with the prior method and how are the DDES simulations used in this procedure (only to generate lift, drag, and moment curves with adequate MCSP?)
technical corrections intro "the chord is suffiecient to study the onset of dynamic stall" > typo
 intro "above 2 M Brunner" inconsistent spacing btwn number and M
 intro "However, installing according devices in a WT" > I think you mean "such devices"
 "investigating dynamic stall on the FFAW3211 airfoil (Fig. 6). The FFAW3211 airfoil is a commonly used geometry for WTs with a heighttochord ratio of 21.1% and the coordinates of the geometry can be found in Bertagnolio" > would rephrase as "investigating dynamic stall on the FFAW3211 airfoil (Fig. 6). The FFAW3211 airfoil is commonly used for WTs with a thicknesstochord ratio of 21.1% and the coordinates of the airfoil can be found in Bertagnolio" (< I made a few changes here)
 line 93: "50×c and a length of 1×c" the crosses are only confusing here, should be removed > 50c and c
 line 110 "(Issa, 1986)" is not a citation for the PIMPLE algorithm, so its presence here is confusing
 line 187: the text write nu be the equation $\nu$
 Fig. 2 should already be given where the model is initially described and not here in the grid study (in any case, no refined mesh is shown) "The error bars are the sampling errors calculated as in Ries et al. (2018).": the reader should not be expected to have read that paper and a few words should be given to explain how the error bars were calculated
 a nonbreaking space is missing for two terms around line 215 (42.7 ... Tab.)
 the whole document uses the notation 3e^4 which in this context means Euler's number, but it is quite clear the authors mean 3x10^4, please use officially recognized notation
 Tab. 4: the e+00 terms are superfluous; also, why is the order 5.07 ?
 line 225: is delta(t) in seconds?
 line 253: "which leads to.." > how does this lead to...? Explain the reasoning. Also: "airfoils are visualized in 6"> missing "figure" word there
 line 264: "The LESP increases quadratically during" you should mention that you are now writing about figure 5 b.
Citation: https://doi.org/10.5194/wes2023159RC2 
RC3: 'Comment on wes2023159', Anonymous Referee #3, 18 Dec 2023
Review of : Predicting the Onset of Dynamic Stall on Large Wind Turbines
By: Jan Dominik Ahrens et al.The paper describe the CFD computation and analysis of a dynamic stall configuration of the FFAW3211 airfoil at a Reynolds number of 10 million, a mean angle of attack of 20 degrees, 5 degree amplitude, and a reduced frequency of 0.137. Before performing the simulation, a grid dependency study is performed for the mean flow angle, to document the necessary resolution. Finally, the computed lift hysteresis is compared to simulations referred to as BEM simulation. Following this, the data is analyzed with respect to shear layer height, and timings with respect to development of dynamic stall vortex and LeadingEdge SuctionParameter and MidChord SuctionParameter.
Detailed comments:
For the present reviewer there is a major problem with the present work, namely the fact that everything is based on a simulation of three cycles of dynamic stall with a DDES model, of which the first cycle is ignored due to initial transients. There is no experimental data for comparison with the performed computations, and the comparison with the BEM setup is not explained in any detail. With respect to the CFD setup, the problem with having only 3 cycles of dynamic stall, is supported by the reference given below:Experimental/Computational Study to Identify Sources of Clusters in Pitching Airfoil Measurements, Steven A. Tran ORCID Icon, Manikandan Ramasamy and Jayanarayanan Sitaraman. Published Online:13 Nov 2022https://doi.org/10.2514/1.J061790
The title is not correct, the present study deals with dynamic stall for a pitching airfoil. It might be argued that the situation mimics what could be experienced on a large wind turbine, but nothing in the study deals with dynamic stall on a large wind turbine.
In the abstract it is claimed that the article focus on development of reduced order models. The present reviewer do not agree upon that reduced order models are developed. A hypothesis is proposed but not proven through the present study. The present analysis is based on a single DDES simulation of a dynamic stall scenario, and the proof of the validity of the proposed MCSP is not provided and would require much further work. Additionally, the present method needs to be implemented in e.g. a vortex code to be useful to predict dynamic stall.
It is a confusing that the authors refer to a BEM method. Normally, one would expect this acronym to cover a Blade Element Momentum theory, which is a theory for rotor aerodynamics not directly relevant for dynamic stall of an airfoil. Most often, a BEM code will have included a dynamic stall model, but a BEM is not a dynamic stall model .
P5, L116: BEM is not normally referred to as reduced order method, but often low fidelity or engineering method.
P2, L57: BEM is not a dynamic stall model, but can easily run at timesteps relevant to dynamic stall.
Dynamic stall P2: L59, I would like the author to point me to the place where Branlard et. al states that ' the Blade Element Momentum (BEM) method only predict the dynamic loads and cannot predict the time scales of dynamic stall'
The introduction of BEM in the article is unnecessary as the present work is not dealing with rotor aerodynamics. One could talk about Blade Element, or thin airfoil theory or dynamic stall models, but the BEM theory should be removed. If BEM is used, we would need details about the turbine setup.
P3:L70: An unsteady vortexlattice method would not normally be described as a reduced order model.
P13,L260: The discrete vortex method of Ramesh et al, would not normally be referred to as a reduced order model.
CFD setup:
The CFD setup is not very typical for DES/LES type simulation. The time discretization is a blend of 1 and 2 order accurate, and the convective scheme is a bounded second order accurate upwind scheme. Normally, one would expect second order accuracy for the time algorithm, and some kind of central difference for the convective terms.It is not clear to the present reviewer if the approach uses a iterative timestepping, and if so what convergence criteria is applied. On page 10 L.225 a timestep of 1.e5 without any dimension (nondimensional timestep ?) is given, which is said to be equal to 1e5 Hz (here with dimensions). The influence of the timestep is not investigated.
The details given about the grid generation is insufficient and do not allow others to reproduce the details. For a DES type simulation, the off surface grid in the separated region is quite important. The zoom around the leading and trailing edge do not allow any judgment of the grid in the DES region. As pointed out by one of the other reviewers, there are huge aspect ratio jumps in the grid at the trailing edge. Additionally, It is not obvious how the refinement is done, as it is not merely a simple doubling in all directions. To judge the validity of the refinement study, one would need to know how the refinement was done.
P4, L94: The justification to use one chord of span by referring to the work of Yaclin is problematic. In the referenced article they are using periodic conditions and not symmetry conditions as done in the present work. There is a huge risk that the effect of the symmetry on the turbulence will be very different from the effect of the periodic condition. Additionally, the symmetry condition will be much more restrictive on the wake development, which could easily for a more 2D flow behavior.
It is surprising that the authors do not at all comment on the predicted 5th order behavior of the method, based on the described method we should expect at most a second order accurate behavior.
Data analysis:
In the results section, very little material is presented. There are no visualizations of the flow, to document the vortex shedding. The practical evaluation of the shear layer height is not discussed. On P11,L240 it is stated that the shear layer is located at the location where U=0.99U_infty. This would normally be the boundary layer height. How is this evaluated for the present dynamic flow setup using DDES. As we do only have two cycles, is it averaged along the span, and is the shear layer height done by visual inspection. It is very difficult to expect this to be very accurate.P14. t*_ss =12.45 but later refer to as t*_SS=12.35, some of it should probably be just t*.
P14, it might not be so striking that the pressure distribution reflects the vortex formations as the pressure and vorticity are heavily coupled. The circulation distribution is changing around the airfoil when the vortex is formed and when it leaves the airfoil.The MidChord SuctionParameter MCSP introduced in the article is calculated by integrating the lift between two rather arbitrary points. The present reviewer expects that this would need to be adjusted for different airfoils, and we need more proof of the generality of the present approach. The authors claim without any proof, that this criteria is more robust than the LESP criteria.
Citation: https://doi.org/10.5194/wes2023159RC3
Status: closed

RC1: 'Comment on wes2023159', Anonymous Referee #1, 06 Dec 2023
The manuscript addresses the need for a reducedorder model to predict the onset of dynamic stall appearing on the wind turbine (WT) blades which inherently have thick airfoil profiles. The authors state that the effective way of delaying the dynamic stall is to develop new airfoil profiles that maximize the angle of attack initiating the stall vortex formation, which is possible by predicting the time scales of dynamic stall stages via reducedorder models. This constitutes the motivation of the study.
To monitor the dynamic stall stages on thick airfoils, the authors propose a MidChord SuctionParameter (MCSP), derived from the LeadingEdge SuctionParameter (LESP) which was previously introduced in the literature and already used for thin airfoils. In the paper, a pitching FFAW3211 airfoil is simulated by DDES, and both LESP and MCSP are tested using the acquired unsteady data. The paper concludes that both parameters are helpful for predicting the onset of dynamic stall stages whereas MCSP seems more reliable as the flow separation occurs at midchord on the thick airfoils.
The study addresses the relevant topics of the journal of WES, and the proposed method (MCSP) might be of widespread interest to the wind energy community and the WES readers. The manuscript, in general, is clear and wellorganized in terms of structure and presenting the results. The literature survey is quite sufficient. However, with the presented results and discussion, it is difficult to come to a conclusion that the MCSP is a more robust criterion for WT airfoils than LESP. Including some additional results that show flow visualizations investigating the flow separation regions and so as the dynamic stall would be beneficial to support the conclusions about the proposed method.
Please find some questions and suggestions to the manuscript as follows:
 Dynamic stall is difficult to capture even with robust CFD techniques. DDES is an accurate and effective tool for massively separated flows where large eddies dominate the field; however, its inherent defects are likely to appear in case of smallscale eddies generated inside the shear layers. DDES models all structures inside the boundary layer until the flow detachment. Thus, the use of DDES may adversely affect the prediction of primary instability and vortex formation stages. Instead, LES or modified versions of DDES (such as IDDES, ZDES, DDESSLA) would be more appropriate (and of course more costly) for this problem. Could you please give reasons for selecting the DDES method with relevant references?
 In Section 2.1, there is a sentence between Lines 105107 stating that the kw SST model provides turbulence closure for 3D RANS in the attached flow regions of DDES. “In the attached flow” phrase might be misleading here because indeed the same equations are solved everywhere in the domain, but an additional switching mechanism in the dissipation term of the kequation makes DDES behave differently in the regions apart from the attached flow. So, I suggest simply removing the phrase “in the attached flow” from the sentence.
 Please clearly define alpha_* and alpha_DS before specifying them, like for alpha_SS.
 In Section 3.2, it is underlined several times that for the investigated case flow is attached near the leading edge and dynamic stall vortex is formed at midchord where the flow detaches. Accordingly, MCSP is calculated using the region of 0.45<x/c<0.55. In addition, (as already mentioned above) MCSP is found by the authors more reliable than LESP due to the flow separation regions. A reader may need to see the flow behavior on the suction side, especially in the LESP and MCSP regions, through some instantaneous flow snapshots. So, it is highly suggested to include additional figures demonstrating the vortex formations and shear layer heights in this section.
 In Section 3.2, the paragraph between lines 232 and 238 discusses the cl results of DDES shown in Fig.4. However, there is no discussion about the BEM results. Please compare DDES and BEM results presented in Fig.4.
 On page 13, the stall delay attributed to the primary instability stage (Deltat_1^*) is compared with literature data using the A/k term of Equation 6. Then, the effect of airfoil geometry on this stall delay is analyzed with the help of Fig.6. In this discussion part, I think there is no need to mention f(beta) as it is not related to the airfoil geometry and also it is not even used in the comparisons. So, I simply suggest removing f(beta) from Equation 6 (also from the text) and directly writing Equation 6 as Deltat_1^*=A/k.
 In Line 277, it says that the MCSP intersects the LESP at t_SS^*. Is there any meaning of this intersection?
 Please find some suggestions/corrections related to grammar, formatting, and typos as follows:
 Line 2: The letter “s” in “simulation” should be capitalized.
 Line 8: No need to abbreviate “LeadingEdge SuctionParameter”.
 Lines 32: “Brunner et al., 2021” should be in parentheses.
 Line 34: “thicknesstochord ratio” instead of “thicknesstochordratio”
 Line 63: What is ROM? Is it BEM?
 Line 68: It should be only “LESP_crit” instead of “critical LESP_crit”.
 Lines 7173: The sentence starting with “LESP_crit is …” may be rewritten as “Since LESP_crit is a function of the airfoil geometry and Reynolds number, using thin airfoil theory it can be predicted with the first term of …”
 Line 77: “stated” instead of “state,”
 Line 90: “thicknesstochord” (or t/c directly) instead of “heighttochord”
 Line 99: “are” instead of “is”
 Line 112: No need to abbreviate “finite volume”.
 Line 137: “CFD” instead of “computational fluid dynamics (CFD)”
 Line 143: Is “alpha” here the induction factor? If so, it should be “a”.
 Line 158: “Model” instead of “model”
 Line 159: “BLM” instead of “BeddoesLeishman Model (BLM)”
 Line 160: “Leishman and Beddoes, 1989” should be in parentheses.
 Line 188: “nu” should be written as its symbol.
 In the caption of Table 4, it should be “ratio” instead of “ration”.
 Line 253: “in Fig. 6” instead of “in 6”
 Line 267: I think instead of “t_SS^*”, “t_DS^*” is intended to be given here, right? If so, it would be better to check “12.35”.
 Line 292: No comma before “demonstrates”.
 Line 296: “thick” is repeated.
 Line 390: The initials of the first author name are missing.
Citation: https://doi.org/10.5194/wes2023159RC1 
RC2: 'Comment on wes2023159', Anonymous Referee #2, 07 Dec 2023
general comments
The submitted preprint proposes the use of a new parameter, namely the MidChord SuctionParameter (MCSP), to better predict dynamic stall on a wind turbine blade when using reduced order models.
The paper starts with a for the most part wellwritten intro giving the reader a good overview of the problem at hand and the state of the art. The subject is also of significance to both industrial and scientific wind energy community and presents a new approach. The figures are well presented and rich in material without also exceeding the capability of the reader to extract information from them. The relevance of the paper is also well justfied in the intro and the formating is generaly very good.
However, the paper leaves the reader somewhat thirsty for more results, comparison and clarification. If the goal of the paper is to convince the community to switch to the MCSP approach, a more thorough study should be conducted. The reader also has the idea the paper was not enough reviewed before being submitted, as there are some inconsistencies, which I relay in the comments below.
The remainder of this review is structured in the "specific comments" and "technical corrections," which each list remarks that the authors should address by expanding or clarifying directly into their paper.
specific comments the names of the boundary conditions are inconsistent between table 1 and figure 1
 it is somewhat confusing that much of the theory presented in the introduction deals with thin airfoils (thin airfoil theory), perhaps because they emanate from the helicopter community, while the paper deals with thick airfoils. A few sentences on the implications of this, regardless of dynamic stall, should be given to the reader
 it remains unclear to the reader without OpenFOAM experience what the boundary conditions actually are: inletOutlet, cyclicAMI, "fixedValue 1e9", nutlowReWallFunction, etc. It is nice to give the OpenFOAMspeccific names for reproducibility, but it should also be explained to the reader what these are in a few words for each of the boundary conditions listed in the table.
 the same goes for Table 2
 although it is not a problem, why was symmetry instead of a periodic boundary condition used on the sides of the airfoil and what are the implications of this?
 the authors mention they developed a new cyclic boundary condition, but no info is given on what is new about it; is cyclic/cyclicAMI not a standard part of OpenFOAM? (although a citation is given, it should not be expected that the reader reads that other paper just to know this and a short sentence should thus clarify the novelty of their sliding interface)
 the mesh at both sides the sliding interface should be shown, especially if the detaching vortices cross it (do they?)
 the text mentions that a constant growth factor is applied for the entire domain, but behind the trailing edge there is at least one jump with approx 2.0 growth, which is not a problem per se but does contradict the text, can you comment on this? Does this come from the relaxation of the normal extrusion parameters (is it hyperbolic extrusion?) or was the mesh adapted by hand (perhaps to project it on the sliding interface?)?
 can you also give delta(z) as z^+ values?
 which software or algorithm was used to generate the mesh?
 what type of shielding model was used for the DDES; with what parameters?
 page 7, lines approx. 163179: this paragraph should be reformulated, it is quite hard to follow and understand what the authors mean
 also, to which coefficient do you refer when your write "normal coefficients"?
 line 188 "freestream velocity" is quite confusing after having two pages of theory which also covers helicopter rotor measurement; you should remind the reader that you are talking about your 2.5D pitching airfoil model. This becomes even more confusing around line 195 where you prevent selfintersection but nondimensionalizing the time with a parameter that depends on U_inf/c (which is a constant, right?). This also needs to be clarified.
 the authors should comment on the impact of having simulated only 3 cycles (of which only 2 are taken for the results) and of the averaging process. It is known that dynamic stall is an process that can vary strongly between cycles and the state of the art is to group the dynamic stall force cycles into multiple clusters, each exhibiting different separation characteristics.
 overall, the introduction and the long description of the BEM and it's (and other reduced order models) lack of proper handling of dynamic stall give the reader the impression that the goal of the LESP and MCSP is to improve such models, but then the LESP and MCSP are obtained from a DDES simulation, and not used for the BEM results that are presented (or if the LESP or MCSP are used to correct the BEM results of figure 4, it is not clearly stated). All this needs to be made clearer in the paper by explaining how the reduced order model (BEM here?) toolchain is modified and what is the difference with the prior method and how are the DDES simulations used in this procedure (only to generate lift, drag, and moment curves with adequate MCSP?)
technical corrections intro "the chord is suffiecient to study the onset of dynamic stall" > typo
 intro "above 2 M Brunner" inconsistent spacing btwn number and M
 intro "However, installing according devices in a WT" > I think you mean "such devices"
 "investigating dynamic stall on the FFAW3211 airfoil (Fig. 6). The FFAW3211 airfoil is a commonly used geometry for WTs with a heighttochord ratio of 21.1% and the coordinates of the geometry can be found in Bertagnolio" > would rephrase as "investigating dynamic stall on the FFAW3211 airfoil (Fig. 6). The FFAW3211 airfoil is commonly used for WTs with a thicknesstochord ratio of 21.1% and the coordinates of the airfoil can be found in Bertagnolio" (< I made a few changes here)
 line 93: "50×c and a length of 1×c" the crosses are only confusing here, should be removed > 50c and c
 line 110 "(Issa, 1986)" is not a citation for the PIMPLE algorithm, so its presence here is confusing
 line 187: the text write nu be the equation $\nu$
 Fig. 2 should already be given where the model is initially described and not here in the grid study (in any case, no refined mesh is shown) "The error bars are the sampling errors calculated as in Ries et al. (2018).": the reader should not be expected to have read that paper and a few words should be given to explain how the error bars were calculated
 a nonbreaking space is missing for two terms around line 215 (42.7 ... Tab.)
 the whole document uses the notation 3e^4 which in this context means Euler's number, but it is quite clear the authors mean 3x10^4, please use officially recognized notation
 Tab. 4: the e+00 terms are superfluous; also, why is the order 5.07 ?
 line 225: is delta(t) in seconds?
 line 253: "which leads to.." > how does this lead to...? Explain the reasoning. Also: "airfoils are visualized in 6"> missing "figure" word there
 line 264: "The LESP increases quadratically during" you should mention that you are now writing about figure 5 b.
Citation: https://doi.org/10.5194/wes2023159RC2 
RC3: 'Comment on wes2023159', Anonymous Referee #3, 18 Dec 2023
Review of : Predicting the Onset of Dynamic Stall on Large Wind Turbines
By: Jan Dominik Ahrens et al.The paper describe the CFD computation and analysis of a dynamic stall configuration of the FFAW3211 airfoil at a Reynolds number of 10 million, a mean angle of attack of 20 degrees, 5 degree amplitude, and a reduced frequency of 0.137. Before performing the simulation, a grid dependency study is performed for the mean flow angle, to document the necessary resolution. Finally, the computed lift hysteresis is compared to simulations referred to as BEM simulation. Following this, the data is analyzed with respect to shear layer height, and timings with respect to development of dynamic stall vortex and LeadingEdge SuctionParameter and MidChord SuctionParameter.
Detailed comments:
For the present reviewer there is a major problem with the present work, namely the fact that everything is based on a simulation of three cycles of dynamic stall with a DDES model, of which the first cycle is ignored due to initial transients. There is no experimental data for comparison with the performed computations, and the comparison with the BEM setup is not explained in any detail. With respect to the CFD setup, the problem with having only 3 cycles of dynamic stall, is supported by the reference given below:Experimental/Computational Study to Identify Sources of Clusters in Pitching Airfoil Measurements, Steven A. Tran ORCID Icon, Manikandan Ramasamy and Jayanarayanan Sitaraman. Published Online:13 Nov 2022https://doi.org/10.2514/1.J061790
The title is not correct, the present study deals with dynamic stall for a pitching airfoil. It might be argued that the situation mimics what could be experienced on a large wind turbine, but nothing in the study deals with dynamic stall on a large wind turbine.
In the abstract it is claimed that the article focus on development of reduced order models. The present reviewer do not agree upon that reduced order models are developed. A hypothesis is proposed but not proven through the present study. The present analysis is based on a single DDES simulation of a dynamic stall scenario, and the proof of the validity of the proposed MCSP is not provided and would require much further work. Additionally, the present method needs to be implemented in e.g. a vortex code to be useful to predict dynamic stall.
It is a confusing that the authors refer to a BEM method. Normally, one would expect this acronym to cover a Blade Element Momentum theory, which is a theory for rotor aerodynamics not directly relevant for dynamic stall of an airfoil. Most often, a BEM code will have included a dynamic stall model, but a BEM is not a dynamic stall model .
P5, L116: BEM is not normally referred to as reduced order method, but often low fidelity or engineering method.
P2, L57: BEM is not a dynamic stall model, but can easily run at timesteps relevant to dynamic stall.
Dynamic stall P2: L59, I would like the author to point me to the place where Branlard et. al states that ' the Blade Element Momentum (BEM) method only predict the dynamic loads and cannot predict the time scales of dynamic stall'
The introduction of BEM in the article is unnecessary as the present work is not dealing with rotor aerodynamics. One could talk about Blade Element, or thin airfoil theory or dynamic stall models, but the BEM theory should be removed. If BEM is used, we would need details about the turbine setup.
P3:L70: An unsteady vortexlattice method would not normally be described as a reduced order model.
P13,L260: The discrete vortex method of Ramesh et al, would not normally be referred to as a reduced order model.
CFD setup:
The CFD setup is not very typical for DES/LES type simulation. The time discretization is a blend of 1 and 2 order accurate, and the convective scheme is a bounded second order accurate upwind scheme. Normally, one would expect second order accuracy for the time algorithm, and some kind of central difference for the convective terms.It is not clear to the present reviewer if the approach uses a iterative timestepping, and if so what convergence criteria is applied. On page 10 L.225 a timestep of 1.e5 without any dimension (nondimensional timestep ?) is given, which is said to be equal to 1e5 Hz (here with dimensions). The influence of the timestep is not investigated.
The details given about the grid generation is insufficient and do not allow others to reproduce the details. For a DES type simulation, the off surface grid in the separated region is quite important. The zoom around the leading and trailing edge do not allow any judgment of the grid in the DES region. As pointed out by one of the other reviewers, there are huge aspect ratio jumps in the grid at the trailing edge. Additionally, It is not obvious how the refinement is done, as it is not merely a simple doubling in all directions. To judge the validity of the refinement study, one would need to know how the refinement was done.
P4, L94: The justification to use one chord of span by referring to the work of Yaclin is problematic. In the referenced article they are using periodic conditions and not symmetry conditions as done in the present work. There is a huge risk that the effect of the symmetry on the turbulence will be very different from the effect of the periodic condition. Additionally, the symmetry condition will be much more restrictive on the wake development, which could easily for a more 2D flow behavior.
It is surprising that the authors do not at all comment on the predicted 5th order behavior of the method, based on the described method we should expect at most a second order accurate behavior.
Data analysis:
In the results section, very little material is presented. There are no visualizations of the flow, to document the vortex shedding. The practical evaluation of the shear layer height is not discussed. On P11,L240 it is stated that the shear layer is located at the location where U=0.99U_infty. This would normally be the boundary layer height. How is this evaluated for the present dynamic flow setup using DDES. As we do only have two cycles, is it averaged along the span, and is the shear layer height done by visual inspection. It is very difficult to expect this to be very accurate.P14. t*_ss =12.45 but later refer to as t*_SS=12.35, some of it should probably be just t*.
P14, it might not be so striking that the pressure distribution reflects the vortex formations as the pressure and vorticity are heavily coupled. The circulation distribution is changing around the airfoil when the vortex is formed and when it leaves the airfoil.The MidChord SuctionParameter MCSP introduced in the article is calculated by integrating the lift between two rather arbitrary points. The present reviewer expects that this would need to be adjusted for different airfoils, and we need more proof of the generality of the present approach. The authors claim without any proof, that this criteria is more robust than the LESP criteria.
Citation: https://doi.org/10.5194/wes2023159RC3
Viewed
HTML  XML  Total  BibTeX  EndNote  

287  114  19  420  19  17 
 HTML: 287
 PDF: 114
 XML: 19
 Total: 420
 BibTeX: 19
 EndNote: 17
Viewed (geographical distribution)
Country  #  Views  % 

Total:  0 
HTML:  0 
PDF:  0 
XML:  0 
 1