xfoil problems

[sechstemann]

Hello guys,

I'm now doing a project in which a FEM and BEM (Blade Element Momentum) coupled simulation is conducted to study the aeroelastic coupling of a large wind turbine blade(NREL 5MW, with hub connection part). I built the model with Python scripting in Abaqus and for the aerodynamic part wrote a script to implement the classical BEM method (not unsteady BEM method). I divided the blade in several sections along the span. In one step of the method i called Xfoil to create the lift and drag coefficients for all the sectional aerofoils which have naturally different Renolds number, Mach number and angle of attack. As input airfoil for Xfoil i take the real coordinates of the profil which don't vary from 0 to 1. My questions are:

1. As input airfoil must the coordinates be normalized or is it also possible to read in real coordinates? Will the result be reliable for creating the coefficients when using the real unnormalized coordinates?

2. Suppose using the real coordinates, then which Renolds number should i specify, rho*vinf*actual chordlength/nu (the actual Renolds number) or rho*vinf*1/nu (the scaled one), which is using the normalized coordinates?

3. The input angle of attack to Xfoil is the angle between the incoming flow and the horizontal axis or the angle between the incoming flow and the chord length direction? say in my case the chordlength, which is reprensented by the coordiantes, doesn't lie in the horizontal direction.

Sorry for typing so long cuz I'm really a new bee with Xfoil. Any help from you guys would be appreciated.

Best Regards,

Yijun

 

[Sparweb]

I don't remember how the input data is formatted in Xfoil so I can't answer #1.
#2 Always use the real Reynolds number for the conditions you are modeling.
#3 Angle of attack is always relative to the chord. That's the definition.

Never apologize for typing too long - you won't get a good answer if you don't ask a good question.

Why are you using Xfoil?
I haven't touched it for a very long time - is it able to give valid results for large angles of attack? You need a wide angle of attack range for WT simulations.
The rest of your method sounds fine. Rotation of the inflow has an important effect that (I think) BEM can't give you as easily as the vortex model does.


PS: The NREL has a forum for getting advice like this. They may be much more helpful that I can be.

https://wind.nrel.gov/forum/wind/index.php?sid=1eae4ac7e957ee8c909eb3e1c0563742

STF

 

[LiftDivergence]

It sounds like you are only using Xfoil for a small part of your analysis - that is, to pull out the infinite (2D) aerodynamic coefficients.

Your coordinates should be "normalized" as you say, so that the input represents the camber lines from zero percent to 100 percent of the chord. XFOIL is very particular about how the coordinates are loaded in, and it does not use the Selig format which has become the standard, if I remember.

When you downloaded XFOIL, it should have come with a sample airfoil which is formatted correctly. It starts at 100 percent, goes over the top, and back around the bottom.

In my experience, if you do intend to use XFOIL for high alphas, it may have problems converging. You can try using the "INIT" command to reinitialize and then run the same AOA again.

There's probably a lot of documentation on XFOIL as it's quite popular. It might be a good idea to see if anybody has tried something similar before, and if there is commentary on the limitations of XFOIL.


Keep em' Flying
//Fight Corrosion!

 

[sechstemann]

Dear SparWeb,

Thx for your reply. After that post I've tested a little bit with Xfoil by using NACA0012. I've tried to load in the original normalized coordinates, the twisted coordinates as well as the upscaled ones and compare their lift and drag coefficients. It seems that from these tests i found:

1. the input alfa in Xfoil should be the angle between the inflow and the horizontal axis

2. with upscaled coordiantes as loaded coordinates, the Re should be downscaled to get the proportinally upscaled Cl and Cd. for example:

NACA0012 normalized Coord NACA0012 scaling factor 2 NACA0012 scaling factor 2
Re = 8e5 Re = 8e5 Re = 4e5
Ma = 0.05 Ma = 0.05 Ma = 0.05
alfa = 7deg alfa = 7deg alfa = 7deg
CL = 0.8208 CL = 1.5933 CL = 1.6414
CD = 0.01162 CD = 0.01941 CD = 0.02324
CM = -0.0078 CM = -0.4049 CM = -0.4387
L/D = 70.63 L/D = 82.07 L/D = 70.62

3. You are right, Xfoil doesn't converge when large AoA is given which is often the case in WT simulation, thus got problems for BEM cuz CL and CD are needed iteratively. What i've implemented in my program now is for these cases i change the AoA a little bit say -0.01 deg and rerun Xfoil commands until it converges.

Somehow I feel it's not a good idea cuz maybe in some steps(altough in the cases i tested so far not yet) alfa will have to be changed too much to get a converged result for Cl and Cd and thus the BEM could be carried on. Even after a little change convergence is already obtained I'm not sure if this pair of Cl and Cd is reliable for the BEM.

In NREL they first get some Cl and Cds for an specified airfoil under a certain AoA range by using wind tunnel tests or softwares. Then they extrapolate the data to get polars for all AoA. Excel tables are also available for several airfoils online. So two ways i figured out to solve the problem: reading Cl and Cd direct from the tables or just write a short program to do extrapolation.

But the problem is that it seems the Renolds number influence on the polars is not considered or at least not mentioned. In BEM i got different Re for every iterative step. Is this effect maybe not significant so that be neglected? Otherwise how should i take this effect into consideration?



Yijun

 

[sechstemann]

Hello LiftDivergence,

Thanks for your reply. After the post I tried something more with Xfoil. It seems that with scaled coordinates Xfoil could also give meaning results if i scale the Renolds number correctly. For instance:

NACA0012 normalized Coord

NACA0012 scaling factor 2​

NACA0012 scaling factor 2​

Re = 8e5​

Re = 8e5​

Re = 4e5​

Ma = 0.05​

Ma = 0.05​

Ma = 0.05​

alfa = 7deg​

alfa = 7deg​

alfa = 7deg​

CL = 0.8208​

CL = 1.5933​

CL = 1.6414​

CD = 0.01162​

CD = 0.01941​

CD = 0.02324​

CM = -0.0078​

CM = -0.4049​

CM = -0.4387​

L/D = 70.63​

L/D = 82.07​

L/D = 70.62​

But i just tried for one airfoil. Must i normalize the coordinates to get reasonable Cl and Cds from Xfoil? In my simulation i read the coordinates from nodes of my FEM model in abaqus.

Under cases of non convergence what i've tried so far is to decrease the AoA a littel bit until it converges. But somehow i don't consider it to be a good idea. You've mentioned the command INIT may help, what does this code do? Reinitialize some conditions for computing or?


Yijun​

 

[Sparweb]

Hi sechstemann,

Looks like you are getting the hang of it.

I do have one more thing to suggest: perhaps you are getting too involved in Xfoil and aerofoil properties when your true goal is a BEM simulation of wind turbine performance? All of the data you are developing has already been done. It is valuable as education, to conduct the calculations yourself, but the essential properties of these airfoils is well known; have been for decades. If you are satisfied with your exploration and understanding of these properties, then you may be able jump over all of the iteration and validation work by simply picking up Michael Selig's work at the UIUC and interpolating if for your load cases.

Good luck to you - it's a lot of fun!

STF