Mlife fatigue analysis of spar buoy hull and tower

[EWEMkite]

Hi all,

I am trying to use Mlife to conduct fatigue analysis of the hull and tower of my floating wind turbine model. As for as I know, I should have at least the nodal loads outpus of both structures. I see that in ElastoDyn it allows maximum 10 nodes to output nodal loads of the tower, but for the hull I am not sure if there is any output channel I could set in either HydroDyn or any other input file. I saw that in HydroDyn it seems to only give options of integrated loadings.

Could anyone give me any advice? Attached is my current ElastoDyn file, in which you may check the output channels I set.

IEA15_CSpar_ElastoDyn.txt
Best Regards.

[jjonkman]

Dear @EWEMkite,

Presumably if you are calculating the fatigue of a spar buoy hull, you are modeling the spar structurally within SubDyn, such that hydroelastic effects and internal member-level loading can be considered within the spar. SubDyn, like ElastoDyn, allows you to output up to 9 reaction loads within each member of the substructure.

Best regards,

[EWEMkite]

Dear @jjonkman,

Thanks for the response. These days I have been reading the documentations of SubDyn. As my model is designed to be made of composite materials rather than traditional steel for both the substructure and the tower, which composes of 1. Inner pultruded panels made of glass fiber in an epoxy resin filament and 2. External filament wound layer made of Silenka Eglass 1200 tex and MY750/HY917/DY063 epoxy, I am not sure how this choice of material will affect my settings in the input files and how would they impact my simulations.

Do you have any suggestion for my simulation based on the composite material choice?

Best Regards

[EWEMkite]

@luwang00 I will try both methods, the problem of using SubDyn is that I worry if the inertia of the ballast could be added correctly, since the cylinder solid is integrated into shell structure. Is there any method for correction？I don't think the inertia would be the same if I set the density of shell to obtain the same ballast mass.

[luwang00]

@EWEMkite I would say, if you don't mind upgrading to v4.1 first, you could give HD filled members a try. I think that also gives you a more realistic description of the internal hydrostatic pressure. I suggest giving the updated documentation on filled members a quick read: https://openfast.readthedocs.io/en/main/source/user/hydrodyn/input_files.html#filled-members. There are significant changes from v4.0.5.

[EWEMkite]

Hi @jjonkman @luwang00,

Sorry to disturb again, I just updated the input files to match the latest openfast version, but I have several questions to confirm before I go ahead:

1. Is the buoyancy computed by strip theory applied at the bottom node of the bottom member? and is the buoyancy distributed along the members through PtfmVol0 of each body if I turn on PropPot? The reason why I switched to strip theory model is that I heard that the buoyancy is applied at the bottom node by strip theory, referring to this paper:
   

2. According to some posts I read, if I use strip theory with PropPot=False for all members, does it mean that though I turn PotMod=1, the .1 .3 and .hst files are actually ignored? I am wondering if it's possible to use .1 and .3 and use strip theory method to get buoyancy rather than using PtfmVol0 of potential flow.

3. I saw that in some posts, potential flow is better for large body, in my case which aims to analyse the member-level loadings, I think strip theory will be the better choice?

Best Regards,

[luwang00]

Hi @EWEMkite,

1. In OpenFAST, the hydrostatic loads or buoyancy on strip-theory members are computed by integrating the hydrostatic pressure over the wetted surface of the displaced structure at each time step. The loads are mapped to the structural nodes in SubDyn accordingly. For a perfectly vertical spar with no tapering as a special case, then, yes, the buoyancy will be all applied to the bottom of the spar and mapped to the lowest node in the SubDyn structural model. In general, it is based on pressure integration, which gives physically correct load distribution (as opposed to based on displaced volume). The buoyance from PtfmVol0 of the potential-flow bodies are separate and added on top of the hydrostatic load on strip-theory members with PropPot set to false.
2. Set PropPot to true for some strip-theory members essentially disables all strip-theory load components on that member except for drag, which is not included in the potential-flow solutions. It only affects the load calculation of that particular strip-theory member and has no bearing on other strip-theory members or any potential-flow bodies included. To mix the potential-flow hydrodynamic loads from .1 and .3 but with strip-theory hydrostatic loads, you need to set PtfmVol0 to zero for all potential-flow bodies, and zero out all entries in the .hst files. For the strip-theory members, you need to set PropPot to false (otherwise hydrostatic load is suppressed) and manually set all strip-theory hydrodynamic coefficients, with the exception of drag coefficients, to zero to avoid double-counting any loads.
3. Yes, for a spar, strip theory is usually good enough because the diameter is much smaller than the wavelength. In HydroDyn, we also have the MacCamy-Fuchs diffraction model for strip-theory members implemented, which further improves the accuracy of the strip-theory model in shorter waves. See https://openfast.readthedocs.io/en/main/source/user/hydrodyn/input_files.html#maccamy-fuchs-diffraction-load-model.

[EWEMkite]

Dear @luwang00,

These comments are very useful for me, thanks for the response!

Best Regards

[EWEMkite]

Hi @luwang00 @jjonkman ,

I tried several configurations of input files:

1. Hybird potential flow+strip theory in HydroDyn, all mass and inertia are included in ElastoDyn, SubDyn was turned off, static equilibrium test was found running well:
   
   Then I ran the same configuration but under an Extreme condition, then the simulation terminated instantly at the beginning:
   
   I tried with finner time step in both MoorDyn and FST, but the simulation still terminated the same. The input files are attached in the first zip. As the simulation didn't use SubDyn, I don't think the bug is in SubDyn.
   openfast(1).zip

2. Hybird potential flow+strip theory in HydroDyn with Ballast modeled with filled member, SubDyn was turned on, only the tower yaw rotation inertia is added in ElastoDyn and all other platform mass and inertia terms are set as zero, static equilibrium test . Similar error trigerred:
   
   I checked the MD.out, it seems that it's numerical unstable, but though I tried with smaller time step it still doesn't work, I think the error is similar to the first situation (extreme condition simulation). This set of input files are in the second zip.
   Static_Equilibrium.zip
   

3. Hybird potential flow+strip theory in HydroDyn with Ballast modeled with filled member, only spar mass and inertia are set in ElastoDyn, disabled SubDyn, static equilibrium. The simulation is running well, but it looks different from the first static equilibrium test (negative surge):
   Static_Equilibrium(1).zip
   

Could you please help me have a check? With the shared openFAST excecution folder in the previous post, you can run my model after changing the directories.

Best Regards,

[EWEMkite]

Dear @jjonkman,

I ran the same configuration of (2) but I can't find the same .out, I think it's because I made some changes today. I checked the .out and found that the tower deflections went unstable. But generally it has the same warning error with the following order:

1. SD_CalcOutput:Small angle assumption violated in SUBROUTINE SmllRotTrans() due to a large Nodal
   rotation.
2. ED_CalcContStateDeriv:SetCoordSy:Small angle assumption violated in SUBROUTINE SmllRotTrans() due to a large tower deflection (ElastoDyn SetCoordSy). and Tower Strike, the simulation then stuck or failed.

I am a bit confused that though I use time step of 0.001 and NumCrctn=4, the simulation still goes unstable instantly:

Today I checked and updated my SubDyn according to the advice in

I have obtained the .json file and put it to the visualizer, but I am not sure how to check the correctness, coud you please give me some guidance?

IEA15_CSpar.SD.CBmodes.json.txt
IEA15_CSpar_SubDyn.txt

Best Regards.

[jjonkman]

Dear @EWEMkite,

From my brief look, your SubDyn modes look OK to me. The Guyan modes are all what I expect them to be. The Craig Bampton modes are first bending (1-2), torsion (3), and second bending (4-5).

I'm curious, does your model run as expected if you consider only the first bending modes, by dropping Nmodes to 2?

Best regards,

[EWEMkite]

Dear @jjonkman,

I have figured out where the bug is.... it turns out that there was something wrong with my HAMS simulation, resulting in wrong hydrodynamic loadings, that was why my model fails even if the wave is small, but it can be run under static equilibrium.

After I replaced .1 and .3 with WAMIT outputs, my model could be run under DLC6.1 stably with Nmodes=4 and DT=0.0025. Higher Nmodes or smaller DT with higher NumCrctn won't work.

However, the computation is still a bit too expensive to me. I tried larger time step with higher NumCrtn, the simulation still quickly goes unstable at the beginning. Though I have used ExctnDisp=0, PtfmYMod=0, WaveDisp=0 in HydroDyn to save computation cost:

I wonder if you have any suggestion to save the computation costs?

Best Regards

[jjonkman]

Dear @EWEMkite,

I'm glad you found the issue and your model is running now! Is this using DT = 0.0025 s and NumCrctn = 1?

To speed up the runtime, have you tried reducing Nmodes in SubDyn to 2, which would likely allow you to use a larger DT? Are there other structural DOFs that you could disable without impacting simulation accuracy that would allow you to use larger DT?

Upgrading to the tight-coupling branch of OpenFAST, which will be released in OpenFAST v5 later this year, may also help to speed up the simulation: #2439.

Best regards,

[EWEMkite]

Dear @jjonkman,

The screenshot uses DT=0.0025, NumCrctn=1, Nmoes=4, and only TeetDOF, DrTrDOF and GenDOF are disabled.

I tried Nmodes=1, but it doesn't help reduce DT. I will try less DoF, I think blade DoFs and TwFADOF2 TwSSDOF2 could be disabled in my model.

I am looking forward to the new version! :)

Best Regards