Dear Jason,
Regarding free decay test for a TLP floating wind turbine, I got surge decay test like the attachment.
I’m wondering if the displacement value is too small because the previous test in the journal indicates the surge starts with -4m to 4m range.
I know it’s not enough data and you can feel it’s difficult to say what is the exact problem. but, with your relative knowledge, Could you explain the reason why I have very small transient value and it is okay to use for my research?
I would really appreciate if you answer my question.
Dear Hailey,
I’m sorry, but I’m not really sure I understand your question. You can set the initial surge displacement from within FAST in order to simulate a surge free-decay test.
Best regards,
I’m so sorry I confused I put my initial displacement with 4m already. I’m done!
Thank you for your reply.
Dear Jason & Dear All,
- Is there any reference value for the damping ratio values from free decay analysis of MIT NREL TLP? I’ve checked Matha (2009) but I can not see.
- I am checking the hydrodynamic solution method (potential flow or strip theory or both) of the MIT/NREL TLP from the “NRELOffshrBsline5MW_MIT_NREL_TLP_HydroDyn.dat” file in the FAST v8 directory. Default value for PropPot=TRUE. Does this mean that only potential flow solution is considered in the solution?
- In order to consider viscous drag forces in a free decay analysis, is it sufficient to add “Additional linear stiffness” & “Additional linear damping” values?
Damping ratio that I get from a surge free decay analysis on the MIT NREL TLP is 2.3% On the other hand, surge period is 63 seconds , which is quite close to the reference value, 60.6 sec.
Thanks in advance for any help.
Sincerely,
Mustafa
Dear Mustafa,
Here are my answers to your questions:
- I’m not aware of a reference values for the damping ratio of the MIT/NREL TLP in various modes of motion.
- When PropPot is set to TRUE in HydroDyn (as for the MIT/NREL TLP model), that means that the fluid inertia, added mass, and buoyancy terms will not be calculated for the strip-theory member; instead, these terms will be calculated in the potential-flow solution. When PropPot = TRUE, only the viscous drag term (not accounted for in potential-flow theory) will be calculated for that strip-theory member.
- You can use the additional linear stiffness and damping if needed, but the viscous terms can be directly calculated through the strip-theory solution.
See the draft HydroDyn User’s Guide and Theory Manual for more information: wind.nrel.gov/nwtc/docs/HydroDyn_Manual.pdf.
Best regards,
Dear All & Dear Jason,
Description: I would like to expand the discussion to MIT NREL TLP analysis with MoorDyn. In the FAST directory, I can’t see the MoorDyn code. Has anyone tried on that before? I understand that MoorDyn is based on a lumped-mass approach which is capable of modelling the mooring dynamics. I also understand that MoorDyn is based on MAP++. I am attaching the codes of the free decay analysis with MoorDyn and the comparison with a free decay experimental test result in sway direction. HydroDyn code is also attached herein.
In summary HydroDYn solution is based on the potential flow theory model, augmenting Strip Theory solution (100 strips), which considers viscous drag effects with the simple drag coefficient. I’ve changed the line stiffness according to the test value.
Result and Questions:
Question 1) Sway response (0-mean). As we see in the comparison, test response is stiffer in sway direction. Damping in the test is also higher than the analysis. At this point do you have any advice for the calibration? In the sway direction, the only stiffness source is the tendons. Two systems (test vs analysis) with same tendon stiffness. For the strips, simple drag coefficient (Cd) value is 0.60.
Question 2) In MoorDyn, I used 0.60 for Cd which accounts for the drag on the tendons which are circular tubes such as the floater. Am I going right with these values? I see that the drag coefficient is dependent on Reynolds number, Carpenter number and surface roughness.
Thanks in advance for your patience with reading and any possible help.
Sincerely,
Mustafa
Codes.rar (5.67 KB)
Dear Mustafa,
The input files for MoorDyn and MAP++ are similar (by design), but the source code is quite different. The MoorDyn source code is included in FAST v8 and OpenFAST.
From your FAST to data comparisons, I see that the FAST model predicts a higher natural period (lower natural frequency) and less damping than the data. The former suggests a difference in stiffness (likely tendon) or mass (physical or hydrodynamic added mass) between the FAST model and experimental model. The latter suggests that the drag coefficient is too small in FAST for this case. I can’t speak to the source or accuracy of your experimental data to recommend a specific calibration approach, but a proper calibration of the FAST model requires an assessment of uncertainty in the experimental measurements (which data is known and to what accuracy, which data is unknown). Model calibration often requires setting up an optimization problem to adjusted inputs in the model with high uncertainty (such as drag coefficients, perhaps stiffness) to find the values that best match data from a set of the calibration cases.
Best regards,
Dear All,
Does keeping CompElast option alone in FEATURE SWITCHES AND FLAGS (of the .fst file) to 1 can set the OpenFAST to perform free decay test?
Initial conditions available are rotor rpm and PtfmPit.
Thanks in advance.
Adding to above, Can anyone let me know, what all changes must be made to keep RNA as a lumped mass
Dear Akheel,
Addressing your first question: In the ElastoDyn input file you can find initial platform displacements (6DOF’s). Give an offset value onetime for each DOF and then can verify how the system decays before it finds a new equilibrium position. Hope this helps
Kind regards
Dear Akheel,
To make the RNA a rigid body, set CompElast = 1 in the OpenFAST primary (*.fst) input file, and disable in ElastoDyn the blade (FlapDOF1=FlapDOF2=EdgeDOF = False), hub (TeetDOF = False, for 2-bladed turbines only), and drivetrain (DrTrDOF = GenDOF = False) DOFs.
Best regards,
Dear All,
I performed free decay simulations for a 10MW OffShore RWT in order to verify the mode frequencies(in particular Pitch and Surge) of the model. The simulations were performed with CompElast, CompHydro and CompMooring active only, where HydroDyn reproduced still water enviroment. For what concern CompElast, I perfomed the decay simulations for completely rigid model, completely flexible model and completely flexible model with the rotor modelled by BeamDyn.
The results obtained for Pitch were consistent according to my opinion, while the results obtained for Surge were not; in particular, the parts of the Surge results that I can’t understand are related to the difference between the damping of rigid model, flexible model and flexible model with rotor modelled by BeamDyn.
In order to be clearer, I specify that the DOFs activated in rigid model are the platform’s DOFs only, the DOFs activated for flexible model are platform’s DOFs+(TwFADOF1+TwSSDOF1) for tower+(FlapDOF+EdgeDOF) for blades and the DOFs activated for flexible model with BeamDyn are platform’s DOFs+(TwFADOF1+TwSSDOF1) for tower and blade’s DOFs associated with BeamDyn modelling.
I attach the results of the free decay analysis, I hope someone will answer about the damping question I’m asking.
Thanks in advance.
Dear Simone,
I agree that the results with BeamDyn look different. I’m not familiar with the model you are running. What solver options have you set in BeamDyn and could numerical damping inherent in the BeamDyn integrator be playing a role? (You could reduce the numerical damping by increasing rhoinf; setting rhoinf = 1 eliminates numerical damping, but you’ll likely need to drop the time step when increasing rhoinf.) What structural damping have you set in BeamDyn? If indeed structural or numerical in BeamDyn is having any effect, it is odd to me that it appears to be reducing the overall level of damping.
Also, does your BeamDyn model give similar results to your ElastoDyn model with only blade DOFs enabled?
Best regards,
Dear Jason,
I performed free decay simulations with flexible rotor and rigid tower without BeamDyn, hence with platformDOFs+(FlapDOF+EdgeDOF) activated, and the results are extremely close to flexible and rigid model for both pitch and surge.
I will leave as attachment the input file I used for the simulations. The structural damping set in BeamDyn individual blade input file is equal to 0.001. I will try to perform the simulation setting rhoinf equal to 1 and I will update with the results.
Thanks for helping.
NREL Forum.7z (15.9 KB)
Dear Jason,
I verified numerical damping question and I can say that numerical dampng is not the source of the problem; I performed the simulation using rhoinf=1 and the results is equal to rhoinf=0.0.
I have no idea about possible cause right now.
Thanks for your help.
Dear Simone,
Do you see any effect when you eliminate the structural damping in BeamDyn?
Best regards,
Dear Jason,
I modified BeamDyn individual blade input file fixing damp_type equal to 0 and the undamped configuration gave same output of the damped one. So I increased the structural damping, first to 0.005 then to 0.01, and the output are still equal to the undamped configuration.
I did not try higher values of damping because they require quite small time step.
I can’t understand why flexible model without BeamDyn is correctly damped and suddenly, modelling the rotor with BeamDyn, the overall damping goes down.
Thanks for helping.
Dear Simone,
I’m glad that the structural and numerical damping are not playing a role…I would not have expected that anyway, but there have been underlying issues with BeamDyn, and so, I want to rule out all options.
I took a brief look at your files, but I don’t see the primary FAST / OpenFAST input (*.fst) file. So, I can’t run this myself. Besides, the HydroDyn file that was attached includes incident waves, which I wouldn’t expect for a free-decay simulation, so, I’m not sure you sent me the correct file(s) anyway.
Regardless, which version of FAST / OpenFAST are you running? There have been many issues in BeamDyn that have been solved over the past few years; when running simulations with BeamDyn, I would recommend upgrading to the newest version of OpenFAST–at this time, the master branch of OpenFAST is v2.4. Does OpenFAST v2.4 give the same results you are showing above?
Best regards,
Dear Jason,
I am sorry about past missing files, I attached two folders, the former includes complete FASTv8 model, the latter includes complete OpenFAST v2.4.0 model(latest release I found).
I followed your advice and I performed free decay simulations using latest model, but I’m in trouble with the update included in the new model.
Even using a rigid model with ElastoDyn only, the simulation stops due to fatal error; the fatal error is related to too large platform’s angle and displacement detected by HydroDyn and ElastoDyn respectively, it is also related to NaN state detected in MoorDyn.
I leave an attachment about the output which is displayed with fatal error.
I think I should probably work on MoorDyn and HydroDyn models in order to solve previous fatal error but I am not so skilled with latest release.
Thanks for your assistance.
NREL Forum-OpenFASTv2.4.0.7z (138 KB)
NREL Forum-FASTv8.7z (139 KB)
Dear Simone,
Looking briefly at your OpenFAST v2.4 model, it appears that the platform mass and inertias in ElastoDyn have been zeroed out. I’m not sure how you converted the model, but this is likely the cause of the numerical instability you are seeing.
Best regards,
