Aerodynamics

Articles on this page:

• Blade aerodynamic loads during idling
• Instability in standstill simulations due to low aerodynamic damping in dynamic stall model
• Inflow angle vs angle of attack
• Polar gradient determination for dynamic stall models using two-point approximation method
• Dynamic stall models produce unexplained spikes or incorrect results
• How to set different aerodynamic properties on each blade in a Bladed simulation
• Polar gradient determination for dynamic stall models using linear fit gradient method
• Warning indicating polar gradient cannot be found using the linear fit method
• Instabilities are observed at a yaw angle between 30-60 degrees in parked or idling
• Determining the starting radius for dynamic stall calculations
• What should the rotational speed be for the Performance Coefficients calculation?
• Where are the Bladed dynamic stall constants from?
• Replacing the dynamic stall module for standstill condition in Bladed
• How to set Vortex Wake parameters
• DWM (Dynamic wake meandering) implementation in Bladed

Problem

How does Bladed calculate loads on the blades in case of idling? Does it utilize BEM method or does it somehow simplify the aerodynamic calculation?

Solution

For idling conditions the standard blade element momentum theory is used, however it is modified by the settings in the “aerodynamics control” screen which defines the TSR range at which zero and full inflow are applied. If the rotor is rotating slowly enough to fall into the ‘zero inflow’ condition, the induction factors are assumed zero and the aerodynamic loads are worked out from the blade geometry, the aerodynamic data tables and includes the effects of dynamic stall.

Keywords

Aerodynamics; Idling

Problem

Why does the BEM aerodynamics yield unreasonable vibrations when operating in stand-still conditions?

Solution

BEM-based aero-elastic models have significant uncertainties when the sectional airfoil sections are exposed to high angles of attack. This is especially true when the unsteady effects are strong, e.g., exposed to a continuous change of the angle of attack due to turbulence. These conditions typically occur in design load case (DLC) of the turbine in standstill, either as parked or as idling in DLC 6.1, DLC 6.2 and DLC 6.3 or in case the turbine shuts down in DLC 1.4. Aerodynamic behavior at high angles of attack is more complex and good experiments and measurement campaigns are difficult to perform. Consequently, there is a wider range in loading predicted by the different models. The following aerodynamic aspects are to be considered when using Bladed for such cases:

1. The IAG dynamic stall model is recommended for simulations under stand-still conditions (parked or idling). The model captures the hysteresis of the aerodynamic forces with good accuracy compared to measured pitching airfoil data at high angles of attack, better than other dynamic stall models in Bladed. Studies show that blade vibrations are smaller when using the IAG dynamic stall model under stand-still conditions at yaw angles of about 30-60 degrees, where instability usually occurs. Please refer to the following articles for more information:
   • Development and Validation of the IAG Dynamic Stall Model in State-Space Representation for Wind Turbine Airfoils
   • Technical modeling challenges for large idling wind turbines
   
   This model is only available from Bladed 4.14 onward (version starting from 4.14.0.3 is suggested). For older versions of Bladed, the recommended practice in point (4) below is suggested.
2. Only the first order stall model is adopted in Bladed, thus higher harmonic effects like airfoil vortex shedding or 3D affects are not modeled.
3. Double check if the number of blade structural parts is sufficient to model flexible blades. Sensitivity studies might be required.
4. If unrealistic vibration still occurs, it is a general practice to increase the structural damping of the first edgewise mode. Note that a sensitivity study might be required to properly scale up the value of the edgewise damping.

Keywords

Aerodynamics; Damping; Idling; Parked; Standstill; Instability

Problem

What is the relationship between inflow angle (phi) and angle of attack (alpha)? What coordinates are they expressed in?

Solution

See Table 7-3 in the Bladed 4.9 user manual, also referring to Figure 7-18.

• Inflow angle: Angle between effective flow speed velocity vector and tangential direction of rotor plane (φ in the figure)
• Angle of attack: Angle between effective flow speed vector and local chord line (α in the figure) expressed at the quarter chord point

The inflow angle is in a coordinate frame based on the un-coned rotor plane, whereas the angle of attack is in the deflected blade station frame.

Thus the cone angle will have an effect on the alpha value, but not the phi value. The only circumstances in which they will be the same is for a blade with no cone, twist or pre-bend / pre-sweep, and with a pitch angle of zero.

Keywords

Angle of attack; Inflow

Problem

How does Bladed determine the gradient for dynamic stall calculations?

Solution

Bladed determines the gradient based on the two distinct values obtained from the lift polar (to be precise normal force polar) provided by the users. One is at an angle of attack equals zero and the other one is at an angle where the normal force becomes zero. This is illustrated in the figure below. A linear slope can then be determined. In case of semi-symmetric airfoil where the zero lift angle of attack is located at angle of attack equals zero, the gradient of the polar exactly at this location is used as the chosen value. From Bladed 4.13, a new method namely "linear fit gradient" approach was introduced and is more robust than the two-point approximation method, see Linear fit gradient method.

Keywords

Angle of attack; Dynamic stall; Gradient; Loads

Problem

Loads are too low or there are unexplained spikes when the dynamic stall models are active but the results are reasonable when the dynamic stall models are deactivated.

Solution

One of the reasons this problem occurs is that the gradient calculations needed for the dynamic stall models are not accurate (see Polar gradient determination for dynamic stall models using two-point approximation method). This occurs most of the time due to gradient changes in the assumed linear region for Bladed to compute the data. As an illustration, see figure below where the gradient changes in the important area between points A and B. Some airfoils indeed have these characteristics, e.g., to introduce an aerodynamic breaking.

For the time being, two solutions can be proposed:

1. If the load analysis is ensured not to deal with these small angles of attack, i.e. calculations never reach these angles, then the polar data in this area can be modified. A simple edit can be done by fitting a linear curve to match the correct polar gradient, then extrapolate the lift data below point B.
2. If the assumption in point (1) does not hold true, then dynamic stall calculations may be deactivated in this blade region.

This issue is fixed from Bladed 4.13 using a newly implemented "linear fit gradient method" described here.

Keywords

Angle of attack; Dynamic stall; Gradient; Loads

Problem

Is it possible to simulate in Bladed with different aerodynamic properties on each blade? How to set up this calculation?

Solution

A workaround is to use ailerons.

Please refer to the demo model and documentation, which can be downloaded using this link. This functionality is enabled via the Project Info screen and the external controller.

Keywords

Aerodynamic; Blade

Problem

How does Bladed determine the gradient for dynamic stall calculations?

Solution

From Bladed 4.13, a new method called the “linear fit gradient method” has been developed and implemented. This method takes into account several polar points to evaluate the real polar gradient by applying a linear fit over the linear polar regime, see Figure below. The gradient of the linear fit approach is calculated by applying the least square approach. The main challenge of applying the linear fit is to employ the method only within the linear regime of the polar. This linear regime is searched automatically from the zero lift angle of attack up to +7 deg. If no reasonable gradient is found within this range, then the gradient is simply set to 2π according to the inviscid gradient from the thin airfoil theory. By using the linear fit approach, the calculated gradient becomes less sensitive to a local change in the gradient value. Using this approach is extremely important when the polar gradient changes between the zero-lift angle of attack and zero angle of attack. By doing so, the dynamic stall calculations will be more accurate in most conditions. This approach is now set as the new default in determining the polar gradient. It is recommended to use this option in Bladed calculations. Illustration about the concept of the linear fit gradient method and examples of its impact for two dynamic stall models in Bladed are presented in the figure below. It can be seen that reasonable results are obtained when the linear fit gradient method is adopted in the computations.

Keywords

Angle of attack; Dynamic stall; Gradient; Loads

Problem

Users may receive a number of warnings when running Bladed simulations such as following:

*** WARNING: Cannot find relevant gradients within polar range for Foil data set 23 to apply the linear fit method. Slope is set to: 6.28319.
*** WARNING: Cannot find relevant gradients within polar range for Foil data set 22 to apply the linear fit method. Slope is set to: 6.28319.
*** WARNING: Cannot find relevant gradients within polar range for Foil data set 23 to apply the linear fit method. Slope is set to: 6.28319.
*** WARNING: Cannot find relevant gradients within polar range for Foil data set 22 to apply the linear fit method. Slope is set to: 6.28319.

Solution

These warning messages are completely fine and safe. This indicates that the polar data at the specific section number provided in the message has a gradient that is outside [+1.8*pi, +2.5*pi] range when it is searched using the linear fit gradient method (see Linear Fit Gradient Method). Therefore a value of 2*pi (more precisely around 6.28318530718) is adopted for the dynamic stall calculations if being defined.

Keywords

Aerodynamics; Stall; Gradient; Polar; Warning

Problem

Why do we observe instabilities at certain yaw misalignment angles during parked or idling conditions?

Solution

The Instability in standstill simulations due to low aerodynamic damping in dynamic stall model article above might explain your issue.

In addition, more detailed analysis of this phenomenon is discussed in the aerodynamics verification report.

This is a long-standing issue that affects aeroelastic simulation tools and is related to the amount of aerodynamic damping/structural damping included in the simulation. In addition, the dynamic stall models are simplified 1st order engineering models designed to capture delays in aerodynamic response. These models are generally validated in 2D test cases. However, the real question is accuracy of the modelling in 3D. In reality, the phenomenon is more complicated where for instance vortex induced vibration may occur that is not captured by these 1st order models typically used in aeroelastic software.

We are pleased to discuss with you case by case if you have this issue as the causes may vary from different blades - please contact us for further support.

Keywords

Aerodynamics; Stall; Yaw; Idling; Parked; Standstill

Problem

Users are allowed to exclude the root area from dynamic stall calculations using StartingRadius parameter. How much should we exclude?

Solution

Dynamic stall is a very complex physical phenomenon which can be modeled using engineering models. The dynamic stall calculations depend on the value of the normal force gradient and the zero lift angle of attack. Especially for the latter, it is not possible to obtain this value from the cylindrical portion where lift is always zero. In fact, dynamic stall only applies to a lifting body and not to a cylinder. In Bladed, the current default value of StartingRadius is 25% of the blade length. However, recent studies found that modern wind turbine blades can adopt lifting sections even to below this criterion. See below figure where the electrical power of the IEA 15 MW wind turbine is still affected when StartingRadius is decreased (upper figure). The load during an IEC wind gust is also still affected (lower figure). Therefore, dynamic stall calculations will be of importance. Users are advised to decrease the value of StartingRadius as small as possible but to exclude the cylindrical sections. Starting from Bladed 4.14, a new method will be adopted where the users do not need to adjust StartingRadius manually. this can be done by simply setting it to zero (will be set as a new default in Bladed 4.14 onwards). Bladed will automatically determine to exclude dynamic stall effects from the cylindrical parts.

Keywords

Aerodynamics; Dynamic stall; Starting radius; Loads

Problem

What should the rotational speed be for the Performance Coefficients calculation?

Solution

If the rotor is rigid, the only thing rotational speed does is to change the Reynolds number, so airfoil properties will also change if they have been defined to interpolate on Reynolds number. In general for a rigid rotor, the performance coefficient results are not very sensitive to the rotor speed.

However, since Bladed 4.7 it has been possible to define flexible blades for this calculation type. Then, the rotational speed does matter, because loads on the blade depend on the rotational speed and that can deflect the blade and change the angle of attack. This leads to a different aerodynamic loads and different performance coefficients.

So, if flexibility is included, it will be useful to run Performance Coefficients calculation with different rotational speeds and find the one that corresponds to the wind speed you are interested in.

Keywords

Performance Coefficient; Rotational speed

Problem

When selecting dynamic stall model, there are some attached flow constants A1, A2, b1, b2. Where do they come from?

Solution

1. All dynamic stall models except IAG model

For all dynamic stall models except IAG model (available since Bladed 4.14), the same default flow constants are adopted:

The flow constants [A1, A2, b1, b2] are adopted from the following work based on the flat plate.

Jones, R, The unsteady lift of a wing of finite aspect ratio, NACA Report NACA-TN-682, 1940.

The pressure lag time constant T_p, the separation time constant T_f, the vortex lift time constant T_v and the vortex travel time constant T_vl are from the following paper (see below table).

Leishman J. State-space model for unsteady airfoil behavior and dynamic stall. In30th Structures, structural dynamics and materials conference 1989 Apr 3 (p. 1319).

2. IAG model

For the flow constants in the IAG model, they are obtained from the default values adopted in the first generation IAG model.

Bangga, G., Lutz, T. and Arnold, M., 2020. An improved second-order dynamic stall model for wind turbine airfoils. Wind Energy Science, 5(3), pp.1037-1058.

The original source of the IAG constants list is as following:

• The attached flow constants [A1, A2, b2] except for b1 are obtained from the Leishman paper, the same as above.
• The pressure lag time constant T_p, the separation time constant T_f, the vortex lift time constant T_v are from the Leishman paper.
• The attached flow constant b1 is tuned based on OSU experimental data for airfoil in the attached flow regime.
• The vortex travel time constant T_vl is tuned based on OSU experimental data for airfoil in the separated flow regime post-stall.

Users are allowed to change these constants if they have reference data to tune them.

Keywords

Dynamics stall; IAG

Problem

Some customers have developed their own dynamic stall algorithms. They want to use their own algorithm to replace the built-in algorithm in Bladed during standstill conditions.

Solution

This can be achieved through the following flowchart:

Keywords

Dynamic stall

Problem

There are some parameters for users to set in the vortex wake features.

Solution

Purpose: capture working starting points for Bladed’s Vortex Wake module. Use short trial runs to balance load accuracy against runtime; adjust per turbine and case.

Azimuth and Time Step

• Set the wake time step from azimuth increment Δθ: f = RPM/60, T = 1/f, Δt = Δθ/360 * T.
• Example: 10 RPM → T = 6 s; Δθ = 3° → Δt ≈ 0.05 s. Faster rotors need smaller Δt.
• Quick Trial Set
• Run 2–3 short cases at Δθ = 3°, 5°, 10°; compare key loads to pick a workable Δθ.

Wake Coverage

• Free wake coverage: start with ~60–75° azimuth. At Δθ = 3°, that is ~20–25 free-wake lines. Increase only if wake evolution detail is important.
• Total wake coverage: 7–10 rotor revolutions usually capture induction. At Δθ = 3°, that is ~840–1200 steps.
• Wake development studies: set “Maximum number of free wake steps” equal to “Max. number of wake steps” (no fixed wake).
• Typical load cases: free-wake steps cover ~60–75°; the remaining steps are fixed wake to reduce cost.

Parameter Quick Reference

• Maximum number of free wake steps: steps where induced velocities are updated.
• Max. number of wake steps: total steps (free + fixed).
• Fixed wake steps: Max – Free; induced velocities not updated.
• Free wake coverage: azimuth span of the free wake.
• Total wake coverage: overall wake span (often 7–10 revolutions).

Adaptive Time Step Period

• Apply a coarser step for the first 10–30 s to accelerate spin-up.
• Evaluate loads only after the wake has matured (e.g., from ~50 s onward).

Cost vs Accuracy

• Smaller Δθ and larger free-wake coverage raise fidelity and cost.
• Practical starting point: Δθ = 3–5°, free wake ~75°, total coverage ~10 revolutions. Relax settings if loads are stable and runtime dominates.

Keywords

Aerodynamics; Free vortex wake

Problem

The customer guide on Dynamic wake meandering feature.

Solution

• Eddy Viscosity uses aeroinfo or steady loads to estimate upstream induction and rebuild the wake deficit.
• Meandering is applied only in time-domain runs, so you cannot export a meandered .wnd.
• You can output the non-meandered deficit as a .wake file, then write your own .wnd.
• Workflow: enable .wake output in Bladed → read deficit → map it into the .wnd grid → optionally scale turbulence/deficit in the wake region → run in Bladed.

Keywords

DWM; Dynamic wake meandering