BeamDyn + UA_Mod=4 in dev-tc requires DT <= 0.001s

[deslaughter]

Bug description

When using UA_Mod=4 with BeamDyn in the dev-tc branch, the time step (DT) needs to be 0.001s or smaller; otherwise, the model will become unstable and the aerodynamic forces will oscillate wildly. The generalized-alpha solver in the tight-coupling algorithm can still make progress, but the runtime increases significantly due to the instability.

To Reproduce

Steps to reproduce the behavior:

1. Compile the dev-tc branch with any compiler.
2. Run the IEA-22MW OpenFAST model with a DT of 0.01s
3. Note the long runtime and frequent Jacobian updates
4. Open output inspect AeroDyn tip node Cl and Cd to find unusual oscillations.

Expected behavior

These nodal values should not oscillate depending on the time step.

Screenshots, if applicable

OpenFAST Version

Build OpenFAST from dev-tc branch.

 **************************************************************************************************
 OpenFAST

 Copyright (C) 2025 National Renewable Energy Laboratory
 Copyright (C) 2025 Envision Energy USA LTD

 This program is licensed under Apache License Version 2.0 and comes with ABSOLUTELY NO WARRANTY.
 See the "LICENSE" file distributed with this software for details.
 **************************************************************************************************

 OpenFAST-v4.0.3-659-gec9b169a-dirty
 Compile Info:
  - Compiler: GCC version 14.2.0
  - Architecture: 64 bit
  - Precision: double
  - OpenMP: No
  - Date: May 14 2025
  - Time: 11:32:48
 Execution Info:
  - Date: 05/15/2025
  - Time: 10:33:05-0400

System Information (please complete the following information):

• OS: macOS 15.4.1
• Compiler: GFortran 14..2.0
• Compiler settings: default

[bjonkman]

I have noticed that the integration method in the continuous UA models sometimes needs a smaller timestep. I assume tight coupling algorithm doesn't include the AD/UA states?

[jjonkman]

@bjonkman -- Correct; the tight coupling only collectively integrates the states of the structural modules (BeamDyn, ElastoDyn, and SubDyn). AeroDyn remains loosely coupled.

[zujama]

Dear @jjonkman, @deslaughter
I am facing this problem right now when using IEA22MW-monopile coupled with Rosco 2.9.0 with DT =0.01, compElasto =1 (instead of 2 which is the original setting). I am trying to simulate 30 different simulations of 4000s, and the simulation sort of freezes at around 1500s with no error. I saw that a possible solution is DT=0.002 as suggested, but this will increase the computational cost and time. I was wondering if there was found another possibility which I did not find in my search.
Note: I used same settings with IEA22MW Semi and did not face any problem, only monopile version that caused this issue. The main difference between the two moduls' settings is that Semi uses UA_Mod=3 while monopile has 4 instead. I tried changing that to three but did not help. Now i understand that the reason is that both are continuous models.
Thank you in advance.
Regard,
Zuhir

[jjonkman]

Dear @zujama,

FYI: If switching from UA_Mod = 4 to 3 does not solve your issue, your issue is different than the one mentioned in the original post.

I'm not an expert on the OpenFAST model of the IEA Wind 22-MW RWT atop a monopile, but the version provided in the IEA Wind repository (https://github.com/IEAWindSystems/IEA-22-280-RWT/tree/main/OpenFAST/IEA-22-280-RWT-Monopile) uses a time step (DT) of 0.0002 s--much smaller than you are using. This model does have BeamDyn enablad, so, disabling BeamDyn should allow for a larger time step, but I'm not aware of what that time step can be. This monopile model does have SubDyn enabled, which can require smaller time steps than models (such as the semi) without SubDyn enabled, and/or a correction step (NumCrctn > 0). I would suggest reducing the time step or adding a correction step to your monopile model.

Best regards,