FAST Blade Input Properties

Good evening,

I have seen in the FAST User Guide that to model the blade mode shapes is possible to introduce two flapwise modes and a edgewise mode in the blade input file. In my particular case, I have data from another edgewise mode and a torsional one and I want to know if it’s possible to introduce them in this file and how to do it.

Thanks in advance,
Marcos

Dear Marcos,

Without customization of the FAST source code, the structural model of FAST v7 and the ElastoDyn module of FAST v8 are limited to two fore-aft and two side-to-side bending-mode DOFs in the tower and to two flapwise and one edgewise bending-mode DOFs per blade. If you’ve looked in the source code, you’ve likely noticed that most things are parameterized into loops through enabled DOFs. As such, it would not be too difficult to customize FAST to allow for additional bending-mode DOFs in the tower and blades. However, we’ve not done this ourselves because as you introduce higher-frequency modes/DOFs, likely modes other than bending will become important (e.g., torsion). To capture effects other than bending requires a change to the beam formulation used for the tower and blades.

For the blades, we’ve introduced a more advanced (nonlinear finite-element-based) beam formulation through the BeamDyn module in the latest release of FAST (v8.12), which among other things, includes virtually unlimited DOFs and torsion. We recommend enabling BeamDyn for blades that require models of higher-fidelity than possible with ElastoDyn.

Best regards,

Thanks for answering Mr Jonkman.

That’s what I needed to know.

Best regards,
Marcos

Good evening,

I’m sorry to bother you again but I’m having problems to define the AeroCent.

Reading the Fast User Guide I have seen that when the blade pitch axis doesn’t pass through the airfoil section at 25% chord we have to use the following equation:

AeroCent = 0.25 - [ (fraction of chord from leading edge to actual pitch axis) - (fraction of chord from leading edge to actual aerodynamic center) ]

In my case, applying this, I have obtained the following results:

Blfract Chord x-t/4 —> Pitch Cent (%) ca (%) AeroCent
0,00 3,30 0,00 —> 0,50 0,50 0,25
0,04 3,41 -0,10 —> 0,50 0,50 0,25
0,08 3,79 -0,19 —> 0,43 0,45 0,27
0,12 4,25 -0,16 —> 0,38 0,40 0,27
0,16 4,57 -0,12 —> 0,36 0,35 0,24
0,20 4,64 -0,09 —> 0,35 0,30 0,20
0,24 4,44 -0,06 —> 0,35 0,25 0,15
0,28 4,15 -0,04 —> 0,36 0,25 0,14
0,32 3,86 -0,03 —> 0,36 0,25 0,14
0,36 3,57 -0,02 —> 0,37 0,25 0,13
0,40 3,31 -0,02 —> 0,38 0,25 0,12
0,44 3,06 -0,02 —> 0,38 0,25 0,12
0,48 2,84 -0,02 —> 0,39 0,25 0,11
0,52 2,64 -0,02 —> 0,41 0,25 0,09
0,56 2,46 -0,02 —> 0,42 0,25 0,08
0,60 2,30 -0,02 —> 0,43 0,25 0,07
0,64 2,16 -0,01 —> 0,45 0,25 0,05
0,68 2,03 -0,01 —> 0,47 0,25 0,03
0,72 1,92 -0,01 —> 0,48 0,25 0,02
0,76 1,81 -0,02 —> 0,50 0,25 0,00
0,80 1,72 -0,07 —> 0,52 0,25 -0,02
0,84 1,63 -0,18 —> 0,54 0,25 -0,04
0,88 1,53 -0,36 —> 0,57 0,25 -0,07
0,90 1,48 -0,46 —> 0,58 0,25 -0,08
0,91 1,41 -0,58 —> 0,60 0,25 -0,10
0,93 1,32 -0,72 —> 0,62 0,25 -0,12
0,95 1,21 -0,88 —> 0,65 0,25 -0,15
0,96 1,04 -1,06 —> 0,71 0,25 -0,21
0,98 0,74 -1,27 —> 0,91 0,25 -0,41
1,00 0,01 -1,51 —> 45,25 0,25 -44,75

These values seem to be ok according to the definition. Near the tip the chord is small and the pitch axis doesn’t pass through the airfoil section, this makes that the fraction of chord from leading edge to actual pitch axis increases its value a lot.

In the other hand I have seen in the Fast User Guide that AeroCent values are limited between 0 and 1. I guess I have misunderstood something. What’s my mistake?

Thank you in advance,
Marcos

Dear Marcos,

I’m not sure I follow your case e.g. what is “x-t/4” and is “ca” the fraction of chord from the leading edge to the actual aerodynamic center? Regardless, I’m assuming you are using FAST v7, from which that definition of AeroCent applies? In FAST v7 and when using AeroDyn v14 in conjunction with FAST v8, it is not possible to model an airfoil that does not through the pitch axis. However, it is possible to model such blade geometry in FAST v8 when using AeroDyn v15, although the specification of the blade geometry is quite a bit different in this case, not based on the old FAST v7 definition of AeroCent.

Best regards,

Good evening Mr Jonkman,

First of all, x-t/4 is the distance in meters between the pitch axis and the aerodynamic center, using the pitch axis as coordinate origin. Then, I have calculated the fraction of chord from leading edge to actual pitch axis as ca(%)-(x-t/4)/chord. The - sign was introduced to be coherent with the coordinate system chosen.

In the other hand, ca (%) is the fraction of chord from leading edge to actual aerodynamic center (sorry, I did not realize it and I used the abbreviation of my native language).

Until now, I’ve tried to implement the model in FAST v.7 using the different examples that I found in this web for the NRELOffshrBsline5MW as reference. If it’s not possible to introduce this kind of geometry, what do you recommend me to minimize the error?

Regardless, I’ll read the AeroDyn v.15 manual to see the differences of the blade specifications. Is there any example created using FAST v.8 and AeroDyn v.15 to download?

Thank you so much for answering.

Best regards,
Marcos

Dear Marcos,

OK, thanks for clarifying.

To model this blade, the easiest thing to do is probably to upgrade to FAST v8 and AeroDyn v15, where you can specify nonstraight blade geometries. If the chord does not pass through the pitch axis near the tip of the blade, my guess is there is some in-plane sweep to your blade geometry that would also effect the structural definition of the blade, in which case the use of the BeamDyn module in FAST v8 to model the blade structural dynamics is recommended.

Test26 in the FAST v8 CertTest is a model of the NREL 5-MW turbine that makes use of BeamDyn and AeroDyn v15. While this model includes a straight blade geometry in BeamDyn, there is an example on our forum where we introduced out-of-plane curvature–see my post dated Jan 28, 2016 in the following forum topic: http://forums.nrel.gov/t/pre-bended-sweep-blade-simulation-with-beamdyn-aerodyn-15/1268/1.

Best regards,

Ok, I’ll do that then. Thanks for everything.

Best regards,
Marcos

Good morning,

I have been following Test 26 to see how to introduce the properties in FAST v8 with ElastoDyn + BeamDyn and AeroDyn 15 and I have two questions.

First of all, in the “beamDyn.dat” file, what’s the meaning of kp_xb, kp_yb and kp_zb? In the post you linked here I have seen that it refers to the curve and the sweep of the blade, but I’m not sure if the reference point for distances is the pitch axis or de mass center. To be sure, are these the distances from pitch axis to each aerodynamic center?

In the other hand, in ElastoDyn file “5MW_Baseline/NRELOffshrBsline5MW_Onshore_ElastoDyn_BDoutputs.dat” I have seen that we have to introduce the blade another time, and it was defined by “NRELOffshrBsline5MW_Blade.dat”. In this case, we have a column called PitchAxis where appears to be introduced, according with the values, the fraction of chord from leading edge to actual pitch axis. In our blade, near the tip, our pitch axis isn’t located in the chord (the pitch axis is located out of the airfoil in all sections near the tip). Even if we virtually extend the chord, this line does not pass through the pitch axis. Then, what distance have to introduce in this case?

Thank you in advance.

Best regards,
Marcos

Dear Marcos,

In BeamDyn, kp_xb, kp_yb, and kp_zb define the curvilinear local blade reference axis (natural geometry of the blade), which locates the origin and orientation of each local coordinate system where the cross-sectional 6x6 stiffness and mass matrices are defined. kp_xb, kp_yb, and kp_zb are defined relative to the blade root and pitch axis. The BeamDyn module does not include any aerodynamic effects, so, the location of the aerodynamic center is irrelevant to BeamDyn; instead the aerodynamic center is specified within AeroDyn v15. We’ve updated the BeamDyn documentation that we will release along with an update to FAST v8 soon (and where we have also renamed “kp_xb, kp_yb, and kp_zb” to “kp_xr, kp_yr, and kp_zr”). Included in this updated documentation is a figure that clarifies the blade reference and local blade coordinate systems, which I’ve also attached to my post dated Feb 18, 2016 in the following forum topic: http://forums.nrel.gov/t/section-coordinate-system-in-beamdyn/1285/1.

When BeamDyn is enabled in FAST v8, the blade-related input specifications of ElastoDyn (including the DOFs, initial conditions, blade input files, etc.) are not used by FAST.

Best regards,

Thank you again Mr Jonkman.

Kind regards,
Marcos

Good evening again,

Following the examples included in the CertTest folder of FASTv8 and with your help I have been able to define our model using BeamDyn and AeroDyn15 but now I’m having convergence issues when I run it.

In particular, I obtain the following message:

“FAST_Solution:FAST_AdvanceStates:BD_GA2:BD_DynamicSolutionGA2:Solution does not converge after the maximum number of iterations
FAST encountered an error at simulation time 4.00000E-3 of 20 seconds.
Simulation error level: FATAL ERROR”

I have tried to solve it increasing the number of iterations first, and the tolerance but the problem remains. Then I tried to compare my model with Test26.fst and I noticed that this one has the same problem activating the platform yaw, roll and pitch tilt rotation degrees of freedom.

In order to check this behaviour and seeing that only this test uses AeroDyn15 and BeamDyn I built a model for both NREL ITI Barge and NREL TLP using the files included in CertTest folder. I chose this ones because they have all the DOF activated. Checking, first of all, that they work perfectly with AeroDyn14 and ElastoDyn I ran the new models using BeamDyn and AeroDyn15 and I found the same convergence problem.

On the other hand, I’ve tried to use your blades in our tower only to be sure that this wasn’t a definition issue and I obtained the same error again.

Faced with this situation I have decided to seek your advice again because the other way I see is return to AeroDyn14 and ElastoDyn and lose the blade curvature in the model.

Thank you in advance.

Best regards,
Marcos

Dear Marcos,

What time step (DT) are you running within FAST and what element order (Order_Elem) have you used in BeamDyn? From my experience in running models with BeamDyn coupled to FAST, these are the two critical inputs. The higher the value of Order_Elem, the smaller the DT will likely need to be. I tend to keep the BeamDyn maximum number of iterations and tolerances as DEFAULT. Before trying to run a BeamDyn model coupled to FAST, I would also first check that the BeamDyn model solves properly and gives reasonable results in standalone BeamDyn (uncoupled from FAST). If you haven’t done that yet, I would start there.

When you enable platform degrees of freedom in Test26, what else have you changed to keep the tower from falling over?

Best regards,

Dear Mr Jonkman,

I used Order_Elem 5 in BeamDyn and DT 0.001 in FAST (I didn’t change this values from Test26). In particular, when I was testing Test26 I thought that fixing the translational DOF could be enough but now seeing it carefully this model don’t have mooring lines and then it would need another constraint.

Thank you for your help. Today I’ll try to run BeamDyn uncoupled to check the results and I will try with other values of Order_Elem and DT.

Best regards,
Marcos

Dear Mr. Jonkman,

Now, decreasing the time step to 0.00125 s both NREL ITI Barge and TLP model run succesfuly with BeamDyn and AeroDyn15.

On the other hand, our blade model doesn’t converge even using BeamDyn uncoupled. I’ve tried to find the differences comparing it with your model and the main one is found in the blade input properties file. I have seen that, for all blade input stations, in the generalized sectional mass matrix was introduced:

m·Ycm = m·Xcm = icp = 0

What’s the coordinates origin for Xcm and Ycm? I wrote this values in our model thinking that it was referred to the pitch axis as origin (in our case is perpendicular to the root and coincident with z axis at this point). Tomorrow I’ll check this values but I don’t find the error.

Finally, I only have data of 30 sections. Could this affect adversely?

Thank you again.

Best regards,
Marcos

Dear Marcos,

As discussed in my post dated Mar 10, 2016 above, the cross-sectional 6x6 stiffness and mass matrices (including X_cm, etc.) are defined in a local coordinate along the (possibly curved) blade reference axis.

The absence of the sectional center-of-mass offsets from the NREL 5-MW turbine in Test26 is discussed briefly in my post dated Feb 23, 2016 in the following forum topic: http://forums.nrel.gov/t/6x6-stiffness-matrix-of-nrel5-mw-turbine/1228/1.

Having a small number of stations where the cross-sectional 6x6 stiffness and mass matrices are defined (input Station_Total) should not adversely effect the BeamDyn solution, unless trapezoidal quadrature is used (Quadrature = 2) and the number of stations, including possible refinement (via input Refine), is less then the order of the shape functions of the finite element (input Order_Elem). 30 stations is not small.

Is the blade geometry you’ve defined via the key points in BeamDyn smooth? If not, BeamDyn can certainly suffer numerical problems. See my posted dated Jan 28, 2016 in the following forum topic for more information: http://forums.nrel.gov/t/pre-bended-sweep-blade-simulation-with-beamdyn-aerodyn-15/1268/1.

Best regards,