FAST errors for high-speed turbulence cases

Hello all,

I am using FAST to simulate the NREL 5 MW turbine (using Test #18 from CertTest as a baseline) under a variety of loading conditions. I am also using Sandia’s 5 MW blade definition (found here: http://prod.sandia.gov/techlib/access-control.cgi/2013/132569.pdf). For the majority of load cases (ECD, EWS, low-speed turbulence) the computations are working well. However, under higher speed turbulence I am getting errors, like the following, early on in the simulation:

 FAST_Solution:FAST_AdvanceStates:AD_UpdateStates:BEMT_UpdateStates(node 9, blade
 3):BEMT_UnCoupledSolve:DeterminePhiBounds:There is no valid value of phi for these operating
 conditions!  Vx = 0.28338, Vy = -20.782, rlocal = 28.159, theta = 0.27689

I understand this to be a fairly general error indicating overall numerical instability. To that end, I have followed the recommendations posted on the topic [url]Initial conditions for tower and rotor] which suggest, at a minimum, having the initial blade pitch and rotor speed appropriately set. The wind case in question is the normal turbulence model (NTM) at a speed of 23 m/s generated by TurbSim. Usually I am using an admittedly coarse grid both spatially and temporally (10x10 grid points and dt = 0.5 s) for the sake of file size and initial testing, but I have also used a finer grid (25x25 grid points with dt = 0.05 s) for this case and it did not seem to help. My initial blade pitch is set to 21.0 degrees and initial rotor speed is 12.25 rpm. Interestingly, if I use the original NREL 5 MW blade properties with initial blade pitch of 0 degrees then the simulation is successful, but the simulation is not successful if I use the original NREL 5 MW blade properties with an initial blade pitch of 21.0 degrees. I believe the blade models should behave similarly with Sandia’s definition being slightly stiffer, although that slight difference seems to be problematic.

I have plotted a few variables from the failed simulation to try to get a better idea of what is going on:
TipDeflection.png
BladePitch.png
RotorSpeed.png
To me, it appears as if the structural simulation is the first to become unstable. That instability is then reflected in the blade pitch control, which is in turn reflected in the overall simulation. I happy to provide more information about this particular simulation to try to diagnose what is going if anybody has any thoughts about additional parameters I should be looking at.

In general, does anybody have any suggestions of additional strategies I could use to avoid these kinds of failures? Perhaps providing initial tower and/or blade deflections would be beneficial for these more extreme cases? I plan to use these simulations for blade optimization procedures, so the reliability of the simulations is fairly important to maintain for a spectrum of blade properties. Perhaps it would be wise to find the steady-state conditions for the original blade design and to use these as initial conditions throughout optimization? I had hoped to not have to go that direction but it wouldn’t surprise me if it ended up being helpful.

Thank you,
Austin Herrema

Dear Austin,

It appears that you model is suffering some sort of numerical instability, but I’m not sure I know enough to identify the root cause.

From my experience, while setting the initial blade pitch and rotor speed are important, setting initial blade and/or tower deflections does not usually mitigate numerical problems. Is there a reason you are initializing the rotor speed to 12.25 rpm instead of 12.1 rpm, which is the rated speed; does initialing the rotor speed at 12.1 rpm yield a better result?

Perhaps the slightly stiffer Sandia blade requires a slightly smaller time step; have you checked the sensitivity of the solution to a lower structural time step?

Best regards,

Dear Dr. Jonkman,

Thank you for your response! I had been initializing the rotor speed to 12.25 rpm because I (perhaps unwisely) used a plot digitizer to extract initial conditions from Figure 9-1 in your original 5 MW reference report (nrel.gov/docs/fy09osti/38060.pdf). Apparently the plot digitizer calibration was not great and I clearly hadn’t thought critically through the actual numerical values–I agree that it doesn’t make sense to initialize the rotor speed beyond rated. To my initial surprise, merely setting the initial rotor speed to 12.1 rpm rather than 12.25 rpm did indeed take care of the problem. Seems like a small difference, but in hindsight I suppose this makes sense as it likely represents an important distinction within the controller. Thank you for your help! It is much appreciated.

Regards,
Austin Herrema

Great, I’m glad that solved the problem! Controllers are often sensitive to the initial conditions.

Best regards,

where can i change the rotor speed?

best regards

Dear Rui.Hao,

The initial rotor speed used in FAST / OpenFAST is set in the ElastoDyn primary input file.

Best regards,

Hi Jason and Austin,
I want to simulate 5MW_OC3Mnpl_DLL_WTurb_WavesIrr with Blade models from Sandia national labs(Sandia’s 5 MW blade definition). Are there any FAST input files for the blades from Sandia that could be integrated into test 19? Any guidance in this regard would be very helpful for me.
Warm regards
Ahmed

Dear Ahmed,

Can you clarify what Sandia blade you are referring to? Are you referring to the following paper: Definition of a 5MW/61.5m wind turbine blade reference model. (Technical Report) | OSTI.GOV? My understanding is this model reverse engineers the original NREL 5-MW blade to derive the composite layup from the original beam properties, but the original beam properties should be similar.

Best regards,

Dear Jason,

yes, I am referring to the same blade you mentioned. I have got the file from the following but I think it is not compatible with openFAST(ElastoDyn).

Dear Ahmed,

The format of the ElastoDyn blade input file has changed a bit from the version you have, but it should not be hard to convert. Here is a summary of the changes:

  • Remove the first header line
  • Remove the line involving CalcBMode
  • Replace the AeroCent column with the PitchAxis column, where PitchAxis = 0.5-AeroCent.
  • Remove all columns after EdgStff

Best regards,

1 Like