I am running into an OpenFAST error - “FF wind array boundaries violated. Grid too small in Z direction” (details in screenshot attached). I am having difficulties interpreting this error. Can you help in providing some insights as to why the wind coordinates are NaN? Does this mean that the wind turbine has collapsed/submerged?
This is a floating wind turbine system with the SubDyn, HydroDyn and MoorDyn modules active. The wind speed is 26 m/s with wind file coordinates of 240 x 270 and the tower height is about 138m, which is within the wind coordinates as discussed in [url]Error with the FF wind array boundaries].
The error is being triggered because the position of an aerodynamic analysis node is NaN (Not a Number). This suggests that your model has gone numerically unstable, resulting in unrealistically large motions. It looks like the simulation failed about 50 s in; I would guess you are seeing a large growth in motion just before the error is reached…is that correct?
In my initial question, I made a mistake with the simulation case. I am running DLC6.1 at EWM 50m/s simulation under differing wind seeds and yaw alignment. When the yaw alignment is -8, I get the error I indicated previously where the aerodynamic node output as NaN. Under a yaw alignment of 0, I get the same error but instead the grid is too small in the Y direction (screenshot attached). Under another wind seed, I get a “member partially submerged” error. Can all these be attributed to numerical instabilities or could there be another reason for these errors?
Focusing on the initial case with the aerodynamic node output as NaN, the tower and platform motion does show some increase (image attached) but not so severe as I was expecting.
Other Simulation notes - I am using the dev branch OpenFAST x64 bits for these simulations. When I use the TCF Mods2 branch x32 bits, I get a “member partially submerged” error for all the DLC6.1 scenarios listed above.
I’m not sure I know what is the underlying problem based on the information provided, but just a few comments:
- The Y-direction error is much more sensible in that the aerodynamic analysis node lies just outside the wind domain, such that increasing the width of the domain may help (assuming the structure is nearing the extent of it motion).
- For the “member partially submerged” error, what is the full error message you are receiving?
- Typically numerical instabilities arise quite early in the simulation. Given that your errors are triggered 50+ s into the simulation, this could mean that instability is more physical in nature (e.g., negative damping of some structural mode that happens at certain conditions).
- I’m not seeing the smoking gun from the images you are providing. You are showing the platform-pitch rate; what does the pitch angle look like? I would also suggest plotting the blade-tip deflections and other platform displacements (sway, heave, roll, pitch, yaw).
- I would not expect any difference between a 32-bit and 64-bit application, other than the 64-bit application can support a higher-resolution model (more analysis nodes); for a model of given fidelity, 32- and 64-bits should be give the same answer.
- The OpenFAST-dev branch has been changing a lot recently, so, it is hard for me to know what is different between that and a TCF mods branch. It would probably be useful to identify specific commits.