Hi everyone, I am modelling a floating wind turbine (12 MW) and trying to compare OpenFAST results with model test (scale 1:60). I need some help in generating the WAMIT data using AQWA. The case I am modeling is a current only case. The numbers and figures that I provide below are all in full scale for ease of understanding.
Below is the 6Dof that I obtained for current only case (0.82m/s), where the surge has a very large period. From model test, the surge is around 4.8 m fluctuating between 4.3-5.3m. Actually the OpenFAST simulation crashed at a bit over 9000 s and fluctuation of 6DOF is getting bigger and bigger. Here only 6Dof is on in ElastoDyn, I am using hybrid model of potential flow and strip theory, WAMIT data is generated in AQWA and members of the floater are given in HydroDyn. If I simulate potential flow only, the second figure is obtained, where the extreme small frequency is obvious and sway and surge are oscillating violently.
In AQWA, I split the floater at water level, and added point mass, cog and moment of inertial. The simulation is done in hydrodynamic diffraction with wave [-180, 180] for every 10 deg, 99 intervals of frequency in [0.002, 0.8] Hz. Mass difference between displaced volume and input mass is around 4%.
I have some doubts:
- I am modeling the wind turbine and floater in model scale by scaling down the dimension, mass and forces with scaling factor, as I am trying to use the same setup as the model test, is there any issue to model it in model scale with OpenFAST?
- To get the added mass, damping matrices and wave excitation forces, is it right just to simulate floater alone (without tower and RNA), give mass of the whole system, COG of the system in the global coordinate (intersection of tower center at water level), and moment of inertial of the floating system including turbine?
- If anyone has experience with conversion from AQWA to FAST, please advice. When I read AQWA to BEMRosseta and save as .hst, .1 and .3 file, I found that all the values in the converted files are different from the output of AQWA, e.g. there is a scaling difference of rho*g in hydrostatic restoring matrix, and rho in added mass. I had to manually scale the values back. Did I miss any step before I save them to WAMIT files in BEMRosseta?
- May I understand it this way for this case, the potential flow mainly contributes to the frequency of the floater, the magnitude of displacement is mainly from the viscous drag from strip theory? If the extreme small frequency 0.000493Hz of surge is not right (the smallest frequency from model test for surge is 0.0116Hz), then the problem is from WAMIT data not strip theory?
- what can be done to eliminate the small frequency and correct the large fluctuation of surge?
Any suggestions and comments are greatly appreciated, I can post the input parameters if required.
Here are my answers to your questions:
Yes, you should be able to model a model-scale floater at model scale, but you may face some numerical challenges associated small time steps and the fact that hard-coded tolerances within OpenFAST are generally set for full-scale models, and so, may not apply at model scale. It is much more common for to model the system at full scale and scale up the tank test data to full scale for comparison.
Are you referring to the AQWA calculation? I’m not familiar with the requirements of AQWA, but the HydroDyn documentation (188.8.131.52. Modeling Considerations — OpenFAST v3.3.0 documentation) provides guidance for WAMIT, where it is recommended to set the vertical CG to zero in WAMIT so that the hydrostatic restoring does not depend on body weight, which is intrinsically accounted for in ElastoDyn, and so, should not be double counted in HydroDyn.
I don’t have any experience with BEMRosetta, but the WAMIT output data is nondimensional. See the following forum post for information on how to dimensionalize the WAMIT output: WAMIT.
While the hydrostatics in potential flow solution always applied, the wave-excitation part of the potential flow solution is not used in a current-only simulation. Exciting the structure via current would happen through the viscous drag term in the strip-theory solution. Once the structure moves, the radiation terms in the potential-flow solution (added mass, radiation damping) would have an effect.
I’m not sure I know enough about your model to comment further.
PS: I notice in your second plot a strong high frequency oscillation in heave, with a nonzero mean. This suggests to me that your model is not in vertical equilibrium, i.e., your full-system weight, undisplaced buoyancy, and vertical mooring pretension do not balance out; I would suggest fixing that per the HydroDyn modeling guidance.
thank you for spending time to read my post and provide helpful suggestions. I have revised my AQWA simulation by setting CG at zero and correcting the mass of the system, and kept the converted output as non-dimensional values. When simulating the case in model scale with current only and new WAMIT input, only 6DOF in ElastoDyn on, CompInflow=0 and CompServo=0, I get below printout on the screen repeatedly, I don’t know if the simulation is running fine or if there is anything that I can do to eliminate this?
Could you please kindly list the hard-coded tolerance? maybe I could change those in the source code.
For now, the reason I stick to model scale modelling is that, the aerodynamic simulation with CL, CD generated with XFOIL for model scale gives an acceptable agreement with model test for thrust and torque, however, for full scale model, I have difficulty to get a set of polar data that can produce good prediction of thrust and torque (~200% overestimation). In Xfoil, I use local Reynolds number (around 2e5) for each section of the blade. For the cases I ran, Ncrit in 1~15 as well as Mach number does not affect CL, CD significantly. For the NREL 5MW and 15MW case, Reynolds number 0.75mil is given, but it is not used in the simulation, to my understanding, and this Re is not the actual one under the operating condition. Under what condition was the polar data generated for these two cases? What Reynolds number shall I use? Actually, when I apply CL and CD for model scale to simulate full scale wind turbine, I get a 10% overestimation of the torque. I am confused about how to choose Re. It would be very much appreciated if you could give some advice.
As discussed several times in this forum, warnings regarding large rotation of the platform suggest that your model is either physically or numerically unstable or susceptible to large motion that cannot be reasonably modeled with OpenFAST. I’m not sure which is the case for your model.
I don’t have a full list of hard-coded tolerances in the OpenFAST source code. I would suggest carefully reviewing all inputs and results to verify their physical correctness.
The Reynold’s number is only used by AeroDyn if you’ve set
AFTabMod = 2. To simulate a model-scale wind turbine at full scale, I would not use
AFTabMod = 2; I would use airfoil data in the full-scale simulation that is valid at the low, model-scale Reynold’s number.