Blade crack modeling in NREL 5-MW baseline wind turbine, modifying EdgeStff

[Borja.Velazquez]

Hello everyone.
I am working on a project about blade crack modeling in OpenFAST. I am modeling it on the NREL 5-MW baseline Onshore wind turbine. For the modeling, it is necessary to reduce the blade stiffness. For this, I am modifying the stiffness only in Edge, that is, only EdgeStff from the Blade.dat file, which is called from ElastoDyn. When I make modifications, I encounter problems in the simulations. The issue I have is that in certain sections out of the 49 in the blade parameterization file, if I change the original value (regardless of whether it is higher, lower, or extreme values), when comparing the final simulation with the healthy blade, the simulations are exactly the same. In other words, it is as if the value had not been changed, and the blade remains healthy.

1. Do you know why this phenomenon occurs? What is the cause?

At the end of the message, I have attached capturas of the blade.dat file, where I modified the EdgStff value on line 27, corresponding to the length 1.495·10^-1 (14.95%). I have problems with this specific line, but not with the next one for example.

I found in this forum that the modification of the EdgStff values must be done simultaneously with the modification of the BLADE MODE SHAPES coefficients. That is, the most realistic modeling involves modifying both EdgeStff and the BLADE MODE SHAPES coefficients. don’t know what the relationship is between the two. I don’t know in what proportion the coefficients change depending on the change in EdgeStff. So here are my questions.

2. What is the relationship between the two?
3. How do I modify these coefficients?
4. Is there any article or document that talks about this topic?

There is a variable, CalcBMode, which does not appear in my Blade.dat file and gives me an error if I add it to my file. If I don’t want to calculate the coefficients:

5. Should I implement this variable?
6. How do I implement it?

Thank you all very much.

Best regards,

Borja Velázquez

Captures off the blade.dat code:

[Jason.Jonkman]

Dear @Borja.Velazquez,

Here are my responses:

1 . The distributed blade mass and stiffness specified in the ElastoDyn blade file get interpolated to the structural analysis nodes specified within ElastoDyn. In the case of the NREL 5-MW baseline wind turbine, there are many more stations in the blade file (NBlInpSt = 49) than there are analysis nodes (BldNodes = 17), so, effectively not all of the data specified in the blade file is used in the interpolation. You could always increase BldNodes (which will slow down the simulation) or reduce NblInpSt to resolve this issue.

2-4. The blade mode shapes are intrinsically tied to the distributed mass and stiffness, so, if you change one, you should change the other. You can use tools such as the old Modes or BModes software provided by NREL or other preprocessors to calculate the mode shapes. Modes and BModes have been discussed in many topics on this forum.

5-6. CalcBMode was proposed at one point, but never implemented. You cannot enable this feature.

Best regards,

[Borja.Velazquez]

Thank you very much for the quick response, @Jason.Jonkman.
If I understand correctly, OpenFAST interpolates those parameters between the BldNodes, right? That is why I am trying to select which NBlInpSt stations on the blade I have to use, whether the most representative ones, all at the same distance, or those that appear in the AeroDyn file. Which do you recommend? I have been trying to choose the most representative ones. And finally, what is the difference between nodes and stations?
Best regards,
Borja

[Jason.Jonkman]

Dear @Borja.Velazquez,

Yes, the data at the NBlInpSt stations are interpolated to the blade analysis nodes. In ElastoDyn, the blade is split into BldNodes equally-spaces elements with the nodes located at the centers of those elements, so, the first node is located a distance of 0.5*(TipRad-HubRad)/BldNodes from the root, the second node is located at a distance of 1.5*(TipRad-HubRad)/BldNodes from the root, etc.

Nodes and stations both refer to locations along the blade, but in this context, stations are where the user inputs distributed blade mass/stiffness data and nodes are were the internal structural analysis takes place.

Best regards,

[Borja.Velazquez]

Dear Jason Jonkman,

Thank you very much again for your quick response. I now have two questions that I am sure you can help me with:

1- What I am doing in my project, as I mentioned earlier, is reducing the Edge stiffness in the section where the crack occurs (with the corresponding changes in the mode shapes coefficients using the BModes tool). However, I am encountering problems and I do not understand the reason. I have been running simulations, and theoretically, the Edge frequency of the blade should decrease if I reduce the stiffness. When performing the simulations, this does not happen. Let me show you some examples, considering the following representations:

• These are the PSD representations of the acceleration at the blade tip. Both the healthy and damaged blades are represented.
• The blue graph is the PSD of the healthy / original blade, and the red graph is the PSD of the damaged blade.
• Each will represent a reduction in stiffness.

The examples are as follows:

• 5% reduction:

• 25% reduction:

• 80% reduction

So, the red graph have to move to lower frequencies, to the left. And the case of 80% have to be lower, no greater than de 25%. I cannot find the error I might be making or if I am missing something. Could you please help me with this issue? Also the frequencies does not match with the outputs of the BModes tool. Is there any particular reason for this?

2- Another question I wanted to ask you is: What is the best output from OpenFAST to observe the edge frequency of the blade? I am currently doing it using the acceleration at the blade tip, and I am not sure if this is the best way.

Thank you very much for everything, your responses and the NREL forum are very helpful. You are doing a great job, Jason.

Best regards

[Jason.Jonkman]

Dear @Borja.Velazquez,

Just a few comments:

• Your PSDs are quite noisy; how do you know that the frequency you are extracting corresponds to the blade-edgewise natural frequency? The quality of the PSD will depend on the simulation length, how you are smoothing the PSD, and how you are exciting the system (I’m not sure a steady 8 m/s wind will even excite the first edgewise natural frequency).
• While natural frequencies play a role, excitation in the rotating frame of references occurs at the rotor speed and its harmonics (1P, 2P, 3P, etc.), which is where I’d expect the PSD to be most dominate.
• Natural frequencies in a rotating frame of reference have to be considered carefully. It typically makes more sense to consider natural frequencies in a fixed frame of reference, i.e., by considering outputs that are in a coordinate system that does not spin with the rotor.
• Deriving natural frequencies by post-processing time series can be difficult as has been discussed in several other forum topics. Instead, I would suggest computing natural frequencies through an OpenFAST linearization analysis, followed by MBC3 and eigenanalysis. NREL has recently introduced a new application (ACDC) that automates the process: GitHub - OpenFAST/acdc: ACDC: Automated Campbell Diagram Code. I would suggest using ACDC to generate and animate the full-system mode shapes of OpenFAST models.

Best regards,

[Daozhi.Tang]

Hi, everyone.
I am working on a project about blade crack modeling too, I think I meet some problems:
1.Do you think my methods is right?
I modified the areo files as following pictures:
Established one ailfoil file which “cd” ,“cm” and “cl” are all zero to simulate fragmented blade(which means that a part of blade drop off)

I assume that blade broken in 26.65m distance from blade root and modified “BlCrvAng”, “BlTwist”, and “BlChord” to 0.001 to avoid solution failed.

I changed the broken blade in Elasto file（”BMassDen”,”FipStiff” and “EdgStiff“ reduce 98%, third row from the bottom )

use BModes to get new mode shapes coefficients.

and I found that the 1st frequency can be matched very well as following picture:
relative error even could be less than 1%, but 2nd not.

2. I don’t know how to explain this. Could you give me some advice?

RtAeroFxh: Total rotor aerodynamic/hydrodynamic and buoyant load (force in x direction)
why there is a vibration happen here？
3. I read the above reply and it is hard to stabilize the system because imbalance of quality from broken blade, so do you have any thought about how to stabilize the system?

[Jason.Jonkman]

Dear @Daozhi.Tang,

Just a few comments:

• You mention the “crack” being at 26.65 m (43% of the 61.5-m long NREL 5-MW blade), but in your AeroDyn blade file, it looks like you’ve modified the properties of the node starting at 18.45 m. And in your ElastoDyn blade file, you change the blade mass starting at 32.8%.
• In your ElastoDyn blade file, I see that you dropped the mass, but not the stiffness. That said, I’d be concerned that a step change to a very small mass and stiffness will not be well resolved by the modal-based method for considering blade structural dynamics in ElastoDyn.
• In your PSD plot, I would expect the strong peaks to be at harmonics of the rotor speed (due to rotational sampling of the turbulent inflow) rather at blade natural frequencies. To assess the change in blade natural frequencies, I would recommend performing a linearization followed by eigenanalaysis of your updated OpenFAST model, e.g. using ACDC.
• I’m not sure I understand enough about our simulation set or other results to comment on your RtAeroFxh results specifically. What do the rotor speed, blade pitch, blade deflections, and tower deflections look like?
• Are you considering just one blade broken, or all three? With only one blade broken, you should see a strong 1P excitation from the rotor imbalance.

Best regards,

[Daozhi.Tang]

Thank you, Jason:@Jason.Jonkman
1.it’s my bad, I type “26.65m“ wrongly , I assumed crack happen on around 32.8% ;
2.yep, I forget to change stiffness; if only use ElastoDyn will not simulate the blade crack appropriately, I will try to use BeamDyn + ElastoDyn method.
3. I think also, please let me update the figure. And I try to perform a linearization followed by eigenanalaysis of your updated OpenFAST model, but failed.
3.I use 8 mps steady wind condtion and simulate only blade 1 broken, Rotspeed:

and blade 1 pitch:

blade 2 tip deflections in x-axis:

and tower top deflections

and PSD(TipDxc2):

and PSD(TipDxc3):

I know why blade 1 tip is not right, because blade 1 tip is not exist in fact but use a very small mass and stiffness.Maybe I should check the new tip deflections after fragment.
Last point about linearization, I only set initial conditon in ElastoDyn files:
12.1 RotSpeed
and .fst switch:

I use CalcSteady to find the stable system, but get :
FAST_Linearize_T:Linearization was forced at simulation end. The linearized model may not be
sufficiently representative of the solution in steady state.

Best regards,

[Jason.Jonkman]

Dear @Daozhi.Tang,

Just a couple comments:

• Your rotor speed seems to continue accelerating, suggesting that the controller is not properly maintaining the speed.
• I’m not sure why your steady-state simulation is not converging. What linearization options are you using? Do the time-series responses during the steady state calculation offer any clue that the steady-state condition will eventually be reached if you increase TMax?

Best regards,

[Daozhi.Tang]

Dear @Jason.Jonkman:
I check my files again and successfully perform a linearization, but if I want perform a linearization system in a stable condition I need to change my DLL to achieve it.
I now find that there ia a difference between BModes‘ results and linearization’s results:
1stFLAP1 is 1.0480Hz by linearization but 0.8399Hz in Bmodes.
1stEDGE1 is 3.2284Hz by linearization but 5.9753Hz in Bmodes.
the following figures are my options:

maybe I should use BeamDyn to simulate broken blades and need use other FEM software to contrast