I am attempting to run OpenFAST, version github.com/HaymanConsulting/OpenFAST.git with a floating wind turbine platform. I have modeled some members as rigid in SubDyn, and also want to perform the Craig-Bampton reductions. Is this possible?
-Dana
Dear Dana,
Glosten reported the same exact error earlier today via e-mail. Are you working together? Emmanuel Branlard is communicating with Glosten about this issue via e-mail. Let’s move this conversation their.
Best regards,
Dear Jason,
I am working in the same topic. My goal is to study the OC4 Semi with SubDyn module on, I’ve create my own input file with the members, joints and sections properties. Apparently it seems to work, however I get the same error as Dana. How can I solve this issue?
Dear Raul,
I would guess there is something incorrectly specified in your SubDyn input file. Are all joints interconnected by members? Do you have any noncantilevered joints (such as pin, universal, or ball) or any members that are not beams (such as rigid links or pretensioned cable elements)?
Best regards,
Dear Jason,
I have checked my SubDyn input file and all joints are interconnected by members, all the joints are cantilever and all the members are beams. Could exist conflict in one joint with two or more members?
Here are somo screenshoots of my input file. All the data I take are from the HydroDyn input file of the OC4 Semi (r-test/NRELOffshrBsline5MW_OC4DeepCwindSemi_HydroDyn.dat at main · OpenFAST/r-test · GitHub).
Thanks for your quick reply!
Dear Raul,
Can you share your complete SubDyn input file?
For the member cross-sectional properties, have you set any mass exceptionally low or any stiffness exceptionally high to mimic rigid or mass-less elements that could cause numerical difficulties with the eigensolver?
I see that you’ve used NDiv = 6, which will subdivide each member, creating many beam elements. Does the problem go away if you reduce NDiv to 1?
Best regards,
Dear Jonkman,
I share you via private message.
Properties are the same from OC4 Semi in HydroDyn, they look logical. I use NDiv=1 and still get the same error.
Best regards,
Raúl.
Dear Raul,
I took a closer look at your SubDyn input file. Joint 2 is the interface joint (connected to the platform reference point in ElastoDyn). You have a member connecting joints 1 and 2, which appears to be a central column below the tower, but no other members connect to these joints. All members should be interconnected and connect to this central column.
Best regards,
Dear Jason,
I’m currently researching the dynamics of flexible semi-submersible floating offshore wind turbines. I’ve been attempting to integrate SubDyn into an OpenFAST simulation of the OC4 semi-sub structure with the NREL 5MW turbine, but I’ve run into persistent issues.
Following the detailed guidance you’ve provided—as well as insights from other experienced users—I’ve ensured that all members and joints are properly connected in both the OC4 configuration and the SubDyn input file as recommended in your comments (The simulation of floating offshore wind turbines considering substructure hydroelasticity · Issue #1250 · OpenFAST/openfast · GitHub) , and this solved SubDyn rigid mode zero eigen issues.
However, I keep encountering a “NaN state detected error “ in MoorDyn. From what I understand, this may be due to structural instabilities.
I also tried your recommended approach (from MD_UpdateStates: NaN state detected - #2 by Jason.Jonkman) of removing MoorDyn from the simulation and compensating by introducing additional stiffness, AddF0, AddCLin, in specific degrees of freedom via the HydroDyn module, but unfortunately, the issue persists.
Would you be able to offer any advice or point me toward a possible solution?
Many thanks,
Paul
Dear @Paul.OsunnaDike,
It sounds like your OpenFAST model is unstable. Have you already tried dropping the time step (DT) or adding a correction step (NumCrctn > 0) in the OpenFAST primary (.fst) input file? You could also try keeping SubDyn enabled, but making it rigid (keeping only the rigid-body Guyan modes), by setting Nmodes = 0 in SubDyn? If your SubDyn model is set up correctly, I would expect its rigid equivalent to match the response of an OpenFAST model where the fully rigid floater is modeled in ElastoDyn rather than SubDyn; does it?
Best regards,
Dear Jason,
Thanks for responding to my query. Following your guidance, I have modified the time step, but the instability persists. I have also run the analysis with SubDyn enabled and keeping only the rigid-body Guyan mode (Nmodes = 0). Comparing the response from this analysis to the case with SubDYn disabled from the .fst file shows different floater responses.
My questions:
- Why is the NaN instability persisting?
- Why does the response with SubDyn (and Nmodes = 0) not match with response with no Subdyn enabled?
- Also, when SubDyn is enabled with rigid-body Guyan modes, I observe brace axial loads of about 20kN . Shouldn’t this be zero or at least numerically insignificant??
I have attached time history plots. The Blue trace is with Subdyn (Guyan rigid-body mode) while the red trace is with Subdyn disabled from the .fst file.
The error file is with Nmodes = 6.
Dear @Paul.OsunnaDike,
Here are my responses:
Regarding (2), when you toggle SubDyn on / off, are you removing the same mass from ElastoDyn that you are adding in SubDyn? (The integrated mass implemented in the SubDyn model is reported in the SubDyn summary file.) Given that your mean heave response is affected by the presence of SubDyn, I would guess you are not doing this.
Regarding (3), internal member-level loads in SubDyn can still be calculated (quasi-statically) when Nmodes = 0 with SttcSolve = TRUE.
Regarding (1), with Nmodes > 0, are you seeing any issues with the Craig-Bampton modes that are kept? FYI: You can visualize the Guyan and Craig-Bampton modes of the SubDyn model through the following website created by Emmanuel Branlard: https://ebranlard.github.io/viz3Danim/. To use this feature, simply load the .json summary file generated by SubDyn.
Best regards,