Dynamic Stall models

Hi all,

Let’s consider a WT aeroelastic model, with rotating beam FEs coupled with a GDW model and complemented with a dynamic stall model for the load calculation (via 2D aerodynamic coefficients).
This model can be represented as a first order dynamics of the kind dx/dt=f(x,t), being “x” a vector containing beam DOFs, their time derivative and induced flow states (something like x=(q; dq/dt; u)).
Nevertheless, in the dynamic stall model, aerodynamic coefficients are also a function of the time derivative of the angle of attack (AOA). Now, being the AOA a function of the induced flow state (u) and of some components of the (dq/dt) (as for instance the velocity of beam nodes perpendicularly to the rotor plane), the final system becomes something like dx/dt=f(dx/dt,x,t).

Is this reasoning correct? Is this the kind of differential system solved within FAST? How can a system like that be treated numerically?

Thanks

Marco

Hi Marco,

You are the correct. A correct way to solve for the coupled structural-aero equations is to combine them in the form dx/dt=f(.,.,t), the structural states forms the inputs to the aero equations as feedback parameters. If the equations are linear, then they would have the form dx/dt=f(x,t). If the equations are nonlinear and have the form dx/dt=f(dx/dt,x,t), then you would need an iteration technique in the time marching scheme.

In FAST, the aero equations are described using the convolution (Duhamel’'s) integrals in discrete-time, approximate solutions to the aero dx/dt=f(x,t). At a time step in the time-marching scheme for the structural equations, the inputs to the aero equations are the current wind input and structural responses from the last time step. The resulting airloads are then used to drive the structural responses.

Cheers,

Khanh

Hello,
It is the first time I’m writing to this forum. We are using the simulation tools FAST and Adams-AdWiMo for wind turbine simulations. Now we found out, that when the wind speed (and with it the angle of attack) is changing rapidly, some operational loads calculated with FAST or Adams-AdWiMo are significantly higher, than using the software Bladed. We think that the problem results from differing dynamic stall models used within Aerodyn and Bladed (Adams-AdWiMo is also using Aerodyn). Both Aerodyn and Bladed use the Beddoes Leishman Model, but it seems, that the implementations are different.
My question is: Do you have done any code comparisons, comparing the dynamic stall models of FAST and Bladed, or do you know any publications? Or do you know something about the deviations?
It would be really helpful to get arguments, that the Aerodyn code predicts the dynamic stall more accurate than Bladed :slight_smile:
I would really appreciate it if you could help me.
Thanks a lot for your time!
Best regards,
Marco

Hello,
We have been doing some comparisons BLADED vs FAST and other tools. I think one source for discrepency is the parametrisation of the stall modell. In FAST you can manipulate a lot the effect by setting the stall-angle in the airfoil file. Normally you estimate the value using the airfoilprep sheet. In Bladed this value is calculated automatically and not really traceable to my knowledge. also Bladed switches the dyn stall modell off in the inner side of the blade by default. This is technically ok but if it is on there in aerodyn you may get for some DLC a significant impact expecially if the blade uses flatback profiles in this region.

Best Regards,
Florian

I just realised that the stall angle parameter in airfoils is no longer used since FAST8.
1 Number of airfoil tables in this file
0.00 Table ID parameter
12.50 Stall angle (deg)
0.00 No longer used, enter zero
0.00 No longer used, enter zero
0.00 No longer used, enter zero
-5.31 Zero lift angle of attack (deg)
6.10840 Cn slope for zero lift (dimensionless)
1.9560 Cn at stall value for positive angle of attack
-0.8000 Cn at stall value for negative angle of attack
-4.5000 Angle of attack for minimum CD (deg)
0.0082 Minimum CD value

Is it now calculated automatically? It was a nice parameter also to quasi disable the dynamic stall in the root region (i.e. to avoid ultra high lift effects).

Here we see the dynamic stall effects during a DLC 1p4 at a inner profile. The blue line is the static lift curve from the airfoil file. The dots are aerodyn outputs:

We see huge lift happening. I think this is not really realistic anymore and that is why in Bladed you disable the Dyn Stall on the inner part of the blade.

Regards,
Florian

FAST 8 and Aerodyn v15 (AD15) use a lot more parameters that you need to set critically for the UA (unsteady aerodynamics) model to work effectively, including alpha1 and alpha2, which are approximately the stall angles for AOA>alpha0 and AOA<alpha0 respectively.

Check out this article for validation data:
dx.doi.org/10.2514/6.2016-1007

If you use Aerodyn v14, there is no difference in the input file when compared to FAST 7 for UA parameters, which means you DO use the stall angle as a proxy for the AOA at which the separation point is at ~70% of the chord. This is more rigorously defined in AD15.

Hello Florian,

thanks for your quick and detailed reply! What you say is really interesting. I have to admit, that my knowledge about dynamic stall is just basic. I just know a bit about the principles.
If I understand you correctly, the dynamic stall model should be switched off in the inner region of the blade.
I thought that especially in the inner region of the blade dynamic stall is occuring, or am I wrong?

Best regards,
Marco

Hello Rick,

thanks for the tip with the article. I’m going to read it soon!

Best Regards,
Marco