The aeroelastic response of a 2 MW NM80 turbine with a rotor diameter of 80 m and interaction phenomena are investigated by the use of a high-fidelity model. A time-accurate unsteady fluid–structure interaction (FSI) coupling is used between a computational fluid dynamics (CFD) code for the aerodynamic response and a multi-body simulation (MBS) code for the structural response. Different CFD models of the same turbine with increasing complexity and technical details are coupled to the same MBS model in order to identify the impact of the different modeling approaches. The influence of the blade and tower flexibility and of the inflow turbulence is analyzed starting from a specific case of the DANAERO experiment, where a comparison with experimental data is given. A wider range of uniform inflow velocities are investigated by the use of a blade element momentum (BEM) aerodynamic model. Lastly a fatigue analysis is performed from load signals in order to identify the most damaging load cycles and the fatigue ratio between the different models, showing that a highly turbulent inflow has a larger impact than flexibility, when low inflow velocities are considered. The results without the injection of turbulence are also discussed and compared to the ones provided by the BEM code AeroDyn.

The current design trend of wind turbines is leading to rotor diameters becoming larger and larger, but they have to be light in order to decrease the cost of wind power generation in terms of leveling energy costs (USD per kWh) and make wind power generation a competitive resource in comparison to other electric generation systems. A lot of research is being carried out to investigate materials and construction techniques in order to allow lighter designs with the consequence that the rotor blades are becoming more and more flexible, which leads to large deformations with associated non-stationary loads and oscillations, resulting in unexpected changes in performances or even flutter if the damping is negative. Additionally, large rotor wind turbines are in reality subjected to diverse inflow conditions, such as shear, turbulence and complex terrain, leading to higher load fluctuations. Moreover, the aeroelastic instabilities strongly affect the operational life of wind turbines

Wind turbines are especially susceptible to fatigue damage, due to the oscillating characteristic of the affecting loads. Fatigue analyses are normally performed by manufacturers for certification purposes, and therefore such analyses are mostly BEM-based. In the EU project AVATAR

Within the scope of the present study, a highly accurate CFD-based aeroelastic model of a 2 MW wind turbine was created and applied to study unsteady load characteristics. The objective was to identify the impact of the modeling of the individual turbine components and the occurring interactions on the transient loads. To achieve this goal, numerical models of successively increasing complexity are introduced. Starting from a one-third model of the blade in uniform inflow, over a complete rotor up to a complete flexible turbine in turbulent inflow, the transient loads were analyzed and compared. The aim was to analyze the main drivers for the load fluctuations and the damage equivalent loading (DEL) using highly accurate models. The different CFD configurations have been analyzed in detail because their computational costs vary enormously. It is therefore of interest, especially for the industry, to know limitations and differences within the high-fidelity approaches. For the uniform inflow case, a comparison with BEM-based calculations is given and two additional inflow conditions are computed, because of its cheapness, in order to determine the generalization level of the results. The ability of BEM in predicting reliable fatigue values changing the computational settings is discussed. In Sect.

The DANAERO wind turbine rotor is used for this paper. This is the reference wind turbine in the IEA Task 29 IV, also known as Mexnext IV

The simulations are performed with the CFD code FLOWer

The CFD model of the blade is created from the provided CAD file, where a “watertight” outer surface is extracted. For the hub, nacelle and tower, surface databases are recreated (cylinder-based) from provided geometrical properties. Meshes are generated by the use of the commercial software Pointwise in combination with in-house scripts. All components have been meshed ensuring

one-third model (BMU) of the rotor (only one blade) suited for uniform inflow conditions,

full model of the turbine (FMU) including nacelle and tower suited for uniform inflow conditions,

full model of the turbine (FMT) including nacelle and tower suited for turbulent inflow conditions.

NAVIER–STOKES and EULER wall represent the ground with and without friction, respectively;

FARFIELD represents the uniform inflow boundary condition;

PERIODIC and PERIODIC ROT represent the symmetrical boundary condition for the full and 120

GUST is the Dirichlet boundary condition, by which arbitrary unsteady inflow can be applied;

PRESSURE OUTLET defines the outflow based on pressure.

Details of the meshes and boundary conditions for BMU

All simulations are run based on the conditions defined in the subtask 3.1 of the IEA Task 29; see

The multi-body dynamics (MBD) simulation code Simpack is used to simulate the structural dynamics of the turbine (as in

Comparison of natural frequencies between the measured ones and those computed by Simpack; single blade above and full turbine below.

Normal (on the left) and tangential (on the right) sectional load in comparison for a single rigid blade 3D CFD vs. AeroDyn.

A simplified aerodynamic model based on Blade Element Momentum (BEM) theory has been generated with the NREL code AeroDyn

Computed cases with uniform inflow in BEM. The first case is the one also computed with CFD.

In order to allow the communication between FLOWer and Simpack, moving, undeformed and reference system markers need to be defined as prescribed in

Explicit coupling strategy.

For the BMU case it was sufficient to run the coupled simulation for only 6 further revolutions to achieve convergence and periodicity of the results. For the FMU, RMU and FMT at least 10 revolutions have been run, although periodicity cannot be reached in the FMT case, because the simulation time is much shorter than the length of the Mann box used. The elapsed time for the coupled simulations (starting from a rigid converged solution) varies from a minimum of 15 h with 1632 processors for the BMU to a maximum of 48 h with 4320 cores for the FMT case. All simulations are run on the SuperMUC-NG supercomputer at the Leibniz-Rechenzentrum in Munich.

All the CFD–MBD computed cases and differences can be seen in Table

Computed cases with inflow condition, CFD modeled structures and flexibility.

The DEL is a constant load that leads, when applied for a defined number of cycles, to the same damage as that caused by a time-varying load over the same period. With this method, two or more signals can be compared in order to obtain insight into the fatigue loadings that blades are facing during normal operation. The approach is based on the S–N curves (stress vs number of cycles) of the material on a log–log scale so that the material behavior is defined by the slope of a line. Additionally, a rainflow algorithm is applied to recognize the relative fatigue cycles in a load signal by filtering peaks and valleys. This algorithm allows us to estimate the amount of load change depending on the amplitude of the cycle. In this way closed stress hysteresis cycles can be identified, defining not only their amplitude but also how often they appear. The consequent damage is, in fact, dependent on the combination of the last two factors. The formulation used in this paper is the one from

In this first section, the effects of aeroelasticity on the reference wind turbine are analyzed. The considered DANAERO experiment was performed at a low inflow velocity (6.1 m s

As validation of the results, the sectional normal (

Comparison of experimental normal to the chord loads (

Results of different field tests have been considered and averaged (black line). As described in Sect.

Tip deformations calculated with CFD at 6.1 m s

The first considerations are made comparing BMU and RMU; the two differ from each other by the presence of a rigid tower and a tilt angle in the CFD model. Deformations in the flapwise, edgewise and torsion direction of the tip of the blade can be seen in Fig.

Thrust and torque calculated with CFD at 6.1 m s

A clear sinusoidal trend can be seen in both cases, which leads to an oscillation of the tip deflection from around 2.3 % to 2.5 % of the blade radius for the BMU case and from around 2.2 % to 2.5 % for the RMU case. The reason for this is the presence of the tilt angle (5

The tip deformations in the edgewise direction are only dependent on the gravitational forces and show therefore almost no difference between BMU and RMU. The same happens for the torsion, whose minimum value is slightly lower in RMU with a very low maximum value of 0.075

Regarding the global thrust and torque in the BMU case for rigid and coupled conditions, it can be seen that

Sectional loads for BMU vs. RMU: comparison between rigid (R) and coupled (C) calculated with CFD at 6.1 m s

The normal forces in coupled and rigid conditions show almost no difference. In the tangential loads, the ones responsible for the power at the shaft, a small increase (around 1 %) can be observed at between 40 % and 60 % of the blade radius, due to a local slightly higher angle of attack (around 0.8 % more), connected with the positive value of torsion shown before and due to the increase in the effective rotor area.

While the CFD calculations have been made based on the operating conditions of the DANAERO experiment, further simulations have been conducted using BEM in order to determine the generalization level of the results. Tip deformations in the flapwise direction can be seen in Fig.

Aeroelastic calculations using BEM as aerodynamic model. Tip deformations in flapwise direction BMU vs. RMU: 6.1 m s

As mentioned in Sect.

Tip deformations calculated with CFD at 6.1 m s

Thrust and torque calculated with CFD at 6.1 m s

Considering the edgewise deflection, the average value increases from 0.43 % of the blade length for RMU to 0.65 % for FMU due to the additional contribution of the tower top deformation. For the same aforementioned reasons, the torsion deflection has on average the same value, but due to the tower's torsion contribution, it shows a higher amplitude of the oscillation that increases in the FMU case by up to 17 % more. The global thrust (

As for the difference in FMU between rigid and coupled conditions, it can be seen that the decay due to the tower passage decreases by 6 % (difference in

Sectional loads for FMU: comparison between rigid (R) and coupled (C) calculated with CFD at 6.1 m s

Pressure distributions for FMU rigid and coupled in comparison calculated with CFD at 6.1 m s

As in Sect.

Aeroelastic calculations using BEM as aerodynamic model. Tip deformations in flapwise direction RMU vs. FMU: 6.1 m s

Figure

Visualization of the

The comparison of the tip deformations in flapwise and edgewise directions and the torsion can be seen in Fig.

Tip deformations comparing FMU vs. FMT calculated with CFD.

The flap and torsion deformations are mostly affected by the presence of turbulence. Especially in the flap direction, five major peaks in 10 revolutions can be observed where the maximum deformation is around 3.1 % of the blade length, which is 47 % higher than the maximum in the uniform case. At the same time, the minimum flapwise displacement, which is not due to the tower passage, is 30 % lower than in the uniform case. For the torsion deformations, the turbulence is mostly affecting the minimum, which for FMU is

This can be explained by the tower top deformations in the flapwise direction in Fig.

Tower top deformation in flapwise direction calculated with CFD.

The spectra of the deformations are depicted in Fig.

Spectra of the deformations comparing FMU vs. FMT.

The effect of the tower can again be recognized in both FMU and FMT with a delay of around 20

The loads resulting from the above-described deformations of the FMT case are shown in Fig.

Global loads in FMT: comparison between rigid (R) and coupled (C) calculated with CFD.

Sectional loads in FMT: comparison between rigid (R) and coupled (C) calculated with CFD.

For the fatigue loading study of the different cases considered, the necessary constants described in Sect.

DEL calculation based on CFD for the different cases using in

Comparison of number of cycle counts to load ranges using

DEL calculation using BEM: results for

The results are shown in Fig.

Switching the FMT case from rigid to flexible increases the DEL, because, as seen in Fig.

Finally, the ability of BEM in predicting the fatigue loading for the BMU and FMU cases is discussed. As can be seen in Fig.

The FMU case is different (no tilt modeling problem occurs), where, for both rigid and coupled and for both chosen input signals, BEM predicts higher fatigue than CFD. The difference between the rigid and coupled case remains the same as that predicted by CFD (so almost none), but the single values are almost 2 times the ones from CFD. The reason for this can be explained by looking at the cycle count in Fig.

In the present work, different computational fluid dynamics (CFD) models ranging from a single blade to the complete turbine including the nacelle and tower of the DANAERO turbine rotor were generated and coupled to a multi-body dynamics (MBD) structural model of the same turbine, by means of a loose (explicit) coupling. The aeroelastic response of the reference turbine was calculated by the use of models increasing their complexity and fidelity in order to recognize differences and deviations connected to modeling approaches in which computational and pre-processing costs strongly differ.
The effects of turbulent inflow conditions were analyzed in comparison to uniform inflow, considering both a rigid and a completely elastic wind turbine model. Additionally, a blade element momentum (BEM) model of the turbine was consistently generated and assessed against the CFD results. In this way it was possible to consider additional uniform inflow cases to determine the generalization level of the results. The objective of this study was to identify the impact and interaction of the different components and modeling approaches on the transient loads and on the damage equivalent loading (DEL) of the blade only. This was evaluated taking into account the flapwise and edgewise blade root moment at the rotor center. The major results of this study can be summarized in the following:

A high-fidelity fluid–structure interaction (FSI) model of the DANAERO wind turbine has been generated and validated in comparison to experimental results.

Modeling the turbine as a single blade instead of entirely leads to only around 1 % to 2 % difference in the average quantities (sectional loads, average torque and deformations). Differently, the resulting DEL increases from BMU (blade only in uniform inflow) to RMU (entire turbine with flexible blades in uniform inflow) by up to 12 times due to the additional large cycles induced by the tower passage and because of the consideration of the tilt angle that leads to a sinusoidal oscillation of the loads, as shown by the BEM calculations.

The introduction of flexibility in BMU increases the DEL because of more load oscillations, which in FMU (entire turbine with uniform inflow) are balanced by a reduction in the tower effect. That is why the DEL was shown to not be affected by flexibility in this case.

When the entire turbine is computed as flexible, a slight increase in the torque is found in comparison to the rigid case at the computed low inflow velocity, due to the increase in the rotor disk area and a slightly increase in the angle of attack (AOA).

BEM shows in general a good agreement with CFD in evaluating the average quantities, although an overestimated tower effect is predicted (with the standard tower model implemented in the AeroDyn version coupled to Simpack) with a direct impact on the DEL evaluation. Additionally, CFD shows a decrease in the tower effect with the introduction of flexibility, which BEM does not show.

Comparing uniform and turbulent inflow, the spectra of the blade tip deformations show that the turbulence increases the amplitude of the broadband while obscuring the higher harmonics of the rotor frequency.

Independently of the rigidity of the turbine, turbulence leads to a much higher amplitude in the load oscillations, in which the tower passage becomes only a neglectable effect. This has a direct impact on the DEL of the blade that increases by up to 11 times in comparison to FMU. Flexibility is indeed additionally increasing the fatigue but much less so than in comparison to what turbulence does, showing that this is the main factor influencing the DEL calculation.

In general it can be concluded that, in the computed cases, turbulence is shown to be the most important factor influencing the DEL of the single blade, more than flexibility, which played in comparison only a marginal role for this specific case where the rotor radius is only 40 m long. Note that when the rotor size increases, the effect of flexibility may play a greater role. Also, the modeling of the turbine as a single blade strongly underestimates the DEL even if CFD is used. On the other hand, a single-blade model (that is much cheaper than a full CFD model of the turbine) is realized to give valid results when just the averaged deformations and loads in uniform inflow are of interest and the predicted tower top deformations are low (as for the low inflow velocity studied in this paper). AeroDyn overestimates the blade–tower effect in comparison to CFD, leading to higher fatigue values, but excluding this overestimated tower effect, BEM can be employed to give useful conclusions regarding the effect of flexibility on fatigue for the uniform inflow conditions under which it has been used.

The raw data of the simulation results can be provided by contacting the corresponding author.

GG generated a part of the CFD model, generated the MBD model, ran the coupled simulations, and performed the post-processing and analysis. GB generated a part of the CFD model. TL and EK supported the research, defined and supervised the work, and revised the manuscript.

The authors declare that they have no conflict of interest.

The authors gratefully acknowledge the DANAERO Consortium for providing the geometry and structural data. They additionally acknowledge Simpack for providing the user licenses and the funders of the project WINSENT (code number 0324129); the Federal Ministry for Economic Affairs and Energy (BMWi); and the Ministry of the Environment, Climate Protection and the Energy Sector Baden-Württemberg under the funding number L75 16012, under which project improvements on the simulation chain were performed. Computer resources were provided by the Gauss Centre for Supercomputing and Leibniz Supercomputing Centre under grant pr94va. Additionally particular thanks are given to the DLR and SWE, University of Stuttgart, for the productive discussions that helped in improving the structural model of these simulations. Finally the authors would like to acknowledge Aimable Uwumukiza for his effort in correcting the language and presentation of this work in its first submission.

This article is part of the special issue “Wind Energy Science Conference 2019”. It is a result of the Wind Energy Science Conference 2019, Cork, Ireland, 17–20 June 2019.

This open-access publication was funded by the University of Stuttgart.

This paper was edited by Katherine Dykes and reviewed by David Verelst and two anonymous referees.