the Creative Commons Attribution 4.0 License.
the Creative Commons Attribution 4.0 License.
| 03 Jun 2026
IEA Wind Task 46 Aerodynamic Benchmark: Computational Aerodynamics Approaches for Assessing Blade Airfoil Performance Reduction due to Leading Edge Degradation
Abstract. Leading edge (LE) surface degradation of wind turbine (WT) blades caused by insect accumulation, erosion and other environmental agents reduces the aerodynamic performance of the blades, causing WT power and energy yield losses. Estimating these losses is paramount for cost-informed maintenance planning. Computational Fluid Dynamics (CFD) can predict aerodynamic performance losses. However, sensitivity of these predictions to physical model choice and detailed model settings can be large. To assess this sensitivity, the International Energy Agency Task 46 – Erosion of Wind Turbine Blades, developed the First Aerodynamic Benchmark, presented herein. The performance degradation of the NACA 633-418 airfoil due to moderate and severe LE degradation, assessed experimentally in two wind tunnel measurement campaigns, is studied. The clean and degraded airfoil performance predicted by seven CFD codes and two low-fidelity methods are cross-compared and benchmarked against measurements. A utility-scale WT featuring the NACA 633-418 airfoil on the outboard blade is used to determine the resulting power and energy losses onshore and offshore. Most codes succeed in predicting the measured airfoil performance reduction due to moderate LE degradation before stall. Consequently, all energy loss estimates are close. Conversely, the variability of the predicted aerodynamic performance reduction due to severe LE degradation is larger, and the variability of the resulting energy losses is also larger than at moderate LE degradation. These results underline both the significant sensitivity to the specific analysis set-up and the need for further research into methods for predicting the impact of advanced LE degradation, such as geometry perturbation-resolving simulations.
- Preprint
(2531 KB) - Metadata XML
- BibTeX
- EndNote
Status: open (until 01 Jul 2026)
- CC1: 'Comment on wes-2026-89', J. Gordon Leishman, 21 Jun 2026 reply
-
RC1: 'Comment on wes-2026-89', Anonymous Referee #1, 23 Jun 2026
reply
Overall Assessment
This manuscript presents a collaborative benchmark study investigating the effects of airfoil-surface roughness and leading-edge degradation on airfoil performance and wind-turbine annual energy production (AEP) using CFD and lower-fidelity methods. The manuscript is particularly valuable as it compares a diverse range of turbulence, transition, and roughness-modeling approaches used by different participants. It evaluates airfoil performance for two roughness heights and subsequently assesses their effects on wind-turbine AEP under both onshore and offshore operating conditions.However, the manuscript has several limitations that should be addressed before publication. Therefore, I recommend major revision. Please refer to the comments below.
Major Comments
1. Presentation and Comparison of the Turbulence and Transition Models
The manuscript should provide a clearer and more concise comparison of the turbulence and transition models used by the different participants. In the current version, it is difficult to identify the differences among the participating groups without repeatedly consulting several paragraphs and the appendices. No consolidated summary of the turbulence-modeling approaches is provided in the main body of the manuscript. I recommend moving Tables A1, A2, and A3, or condensed versions of these tables, into the main body of the manuscript. These tables contain essential information needed to interpret the results. Presenting this information earlier would allow readers to understand the differences among the modeling approaches before examining the comparative results.2. Precise Identification and Citation of the Turbulence and Transition Models
The manuscript should pay more attention to the precise identification and citation of the turbulence and transition models. For example, the SST model is cited using both Menter (1994) and Menter et al. (2003). Although the differences between these SST formulations are relatively limited, the specific model version used by each participant should nevertheless be stated clearly.This issue is more important for the transition model. The manuscript cites Menter et al. (2006), Langtry et al. (2006), and Langtry and Menter (2009) when discussing the transition-model formulation. However, the model formulations and empirical correlations differ considerably among these publications. For example, the earlier formulation included an acceleration parameter, (K), and did not publish the complete set of empirical correlations, whereas the later formulation removed this parameter and provided the complete correlations.
As demonstrated in studies such as the NASA Turbulence Modeling Resource validation cases and the AIAA Transition Modeling Workshops, relatively small differences in model formulations and implementations can produce substantially different predictions. Therefore, it is not sufficient to refer generally to the “(\gamma-Re_{\theta})” transition model. The authors should identify the exact model version and implementation used by each participant.
Considering the already considerable length of the manuscript and the diversity of the numerical approaches, the authors could consider removing or reducing the general explanatory subsections on the SST and transition models (Section 3.1) particularly because not all participants used these models. Instead, the manuscript could provide accurate references and concise descriptions of the exact model versions used by the individual participants.
3. Readability and Presentation of the Figures
The overall readability of many figures is poor, including Figures 2, 4, 5, 7, 8, 11, 12, 14, 15, 17, and 18. It is difficult to distinguish the results of the different participants because of the large number of curves and the limited differentiation among line styles, symbols, and colors. In addition, some of the y-axis ranges are too broad or otherwise unsuitable for identifying meaningful differences among the predictions. The authors should revise these figures by using more distinguishable line styles and symbols and by increasing the font and marker sizes. The axis ranges should also be adjusted where necessary to make differences among the predictions more visible.For Figures 14, 15, 17, and 18, please also provide a caption or legend identifying the dashed line.
4. Differences in Equivalent Sand-Grain Roughness: Tables A2 and A3
The equivalent sand-grain roughness heights differ among the participants, and several participants appear to have used values that were not derived directly from the experiments. Please explain why different equivalent sand-grain roughness heights were selected. The use of different roughness values raises concerns regarding the fairness and interpretation of the benchmark comparison. If the objective is to compare different approaches for roughness modelling, readers would generally expect all participants to use the same equivalent sand-grain roughness height.
The authors should discuss this issue explicitly and clarify why the experimental equivalent sand-grain roughness heights were not used by all participants.Specific and Minor Comments
Please verify the citation format throughout the manuscript, including Lines 35 and 40.
- Line 35
Current text: rain Keegan et al. (2013)
Suggested revision: rain (Keegan et al., 2013)
- Line 40
Current text: LE structure at severe damage stages Han et al. (2018); Im and Kim. (2019); Cappugi et al. (2021); Campobasso et al. (2023).
Suggested revision: LE structure at severe damage stages (Han et al., 2018; Im and Kim, 2019; Cappugi et al., 2021; Campobasso et al., 2023).Lines 129 and 160
The manuscript states:
“The roughness element height is (K=200\ \mu\text{m}), corresponding to a value of (K=246\ \mu\text{m}) for a unitary chord.” and
“The equivalent sand-grain roughness of the P40 sandpaper is (K_s \approx 425\ \mu\text{m}), corresponding to a value of about (708\ \mu\text{m}) for a unitary chord.”
Please also provide the equivalent sand-grain roughness for the lower-roughness case. The discussion currently presented in Lines 469–470 could be moved to immediately after Line 129. The equivalent sand-grain roughness could then be briefly restated in the later discussion for the reader’s convenience.Line 385
The roughness-induced transition model of Langel et al. (2017) was originally developed in conjunction with the Langtry–Menter two-equation transition model. Please briefly explain how this roughness model was implemented in RFOIL and how it was coupled with the transition-prediction method used in the present study.Table A1
Please define “Ambient turb.” more clearly. Does this term indicate the use of the sustaining terms by Spalart and Rumsey (2007)?
The table currently reports the physical first-cell height. Please provide the corresponding nondimensional wall spacing, (y^+), because this quantity is more informative. It is unclear whether the mesh-topology column provides useful differentiation because all participants used structured (all-quadrilateral) meshes. Other mesh parameters may be more informative, including:
• Wall-normal growth ratio
• Leading edge / trailing edge spacings.Citation: https://doi.org/10.5194/wes-2026-89-RC1
Viewed
| HTML | XML | Total | BibTeX | EndNote | |
|---|---|---|---|---|---|
| 81 | 19 | 4 | 104 | 7 | 5 |
- HTML: 81
- PDF: 19
- XML: 4
- Total: 104
- BibTeX: 7
- EndNote: 5
Viewed (geographical distribution)
| Country | # | Views | % |
|---|
| Total: | 0 |
| HTML: | 0 |
| PDF: | 0 |
| XML: | 0 |
- 1
The manuscript addresses a relevant benchmark problem, namely, predicting the aerodynamic performance loss of wind-turbine airfoils caused by leading-edge degradation. However, the paper is unnecessarily long and poorly focused. Its actual contribution is the comparison of several CFD and low-fidelity methods against wind-tunnel data for the NACA 63_3-418 airfoil, followed by propagation of the resulting airfoil polars into turbine power and AEP estimates. That is a useful benchmark exercise, but the manuscript buries it under excessive background material, generic CFD exposition, and lengthy descriptions of standard models.
A major weakness is the inclusion of textbook-level turbulence, transition, and rough-wall-function equations that are neither original nor necessary to interpret the benchmark. Reproducing the $\gamma-Re_\theta$ SST transport equations, SST closure equations, wall-function formulae, and routine solver details does not strengthen the paper. It creates the appearance of technical depth while distracting from the actual scientific issue: whether current modeling practices can reliably predict performance degradation due to leading-edge roughness and erosion. Most of this material belongs either in cited references, a compact methods table, or an appendix.
The paper should be reorganized around the benchmark outcomes rather than generic CFD documentation. The main text should identify the modeling choices that differ among participants, explain which of those choices materially affect the results, and quantify the consequences for lift, drag, stall behavior, power loss, and AEP loss. The authors should also be more critical about the limitations revealed by the benchmark. If the predictions are highly sensitive to roughness modeling, transition treatment, wall functions, turbulence model constants, grid choices, or user-specific setup decisions, then the main conclusion should not be framed as a successful validation of current methods. It should be noted that the current RANS-based roughness modeling remains fragile and only partly predictive, especially for severe leading-edge degradation.
In summary, in its present form, the manuscript reads more like a compilation of participant CFD setup notes than a sharply written benchmark paper. The useful findings are diluted by redundant methodological padding and insufficient critical interpretation. I recommend a major revision. The manuscript should be substantially shortened, with standard equations removed from the main text, solver descriptions compressed, and the discussion refocused on what the benchmark actually demonstrates about predictive capability, uncertainty, and the limitations of current aerodynamic degradation models.