Transient behaviour in FAST

Wind & Water Computer-Aided Engineering Software Tools
1

Dear Dr.Jonkman,

I’ve recently learnt how to use FAST and have been deriving a set of RAOs for the surge motion of the NREL 5MW baseline turbine affixed to the MIT/NREL TLP. I’ve been using the input files that you had provided on your website.

I just have a quick question regarding the transient behaviour at the start of the simulations. Whilst I was running a simulation for a steady wind speed of 20m/s under regular waves of significant height 2m and frequency 0.2rad/s I saw the attached variation of platform surge with time. It seems like there’s a superposition of 2 waves, and convergence still isn’t reached after 4000s. I have both the pitch and variable control switches activated and have been using the user-defined routines available in the DISCON and DFORRT .DLL files you had on your website. I’m not understanding what exactly is going on here, as I do not see this for a wave frequency of 0.1rad/s.

I would appreciate it if you could provide any insight on this. I searched around the forum but couldn’t find similar posts, though I apologise if it’s been asked before and I just didn’t see it.

Thanks in advance

Kurt

Attachment:

Surge.png
Surge.png1409×549 34.5 KB

2

Dear Kurt,

It is hard to interpret the response from the time series alone. I suggest you look at the power-spectral density (PSD) or FFT of the time series to identify what frequencies/modes are playing a role in the response.

Please also realize that FAST is not a linear model, so, simulating with sinusoidal wave inputs does not gaurantee to result in a purely sinusoidal response at the wave period.

Regardless, our preferred method for computing RAOs with FAST (with or without an operational wind turbine) is to run a simulation with banded white-noise wave excitation. This avoids the hassle of running many long simulations waiting for transients to die out and trying to identify the amplitude/phase of the response to periodic wave input.

Best regards,

3

Dear Dr.Jonkman,

first of all, thanks for your reply!

Secondly, I have been slowly working towards generating the RAOs for each wind speed using plane progressive (regular) waves. Here is the method I was following:

  1. Apply wave frequency of 0.1rad/s (spectral period of 62.83s, and regular waves of significant wave height 2m
  2. run FAST for a number of seconds to ensure stabilisation of results
  3. view output time series for platform surge motion (as attached in previous post)
  4. determine at which point the results are stable and then find the peak to peak surge motion and divide by 2 to obtain the response amplitude - this would be the nondimensional RAO value since the applied wave amplitude is 1m
  5. add value to table of results and apply next wave frequency (corresponding to 0.2rad/s) and repeat process

I was doing this for a number of wind speeds and I started from a wind speed of 11m/s to compare with those generated by Matha and Aina Crozier in their theses. From what I could see my resulting RAO graph for that wind speed matched theirs so I assumed my method was correct. Prior to starting I had gone through their methodology and if I’m not mistaken that’s what they did too, granted that I understood what they did (!). What do you think of the above?

I looked into the banded white-noise application but I am using version 7.02 of FAST and have not found anything about it in the manual or in the input files. If my above method is flawed could you kindly guide me as to what I should do?

Thank you very much for your help

best regards

Kurt

4

Dear Kurt,

Your process sounds reasonable, altough it would take many long simulations to get a fine resolution of the RAOs as a function of wave frequency and wind speed, and you’ll run into the problems I mentioned in prior post (nonsinusoidal response from FAST to periodic wave input). The preferred method would be to use banded white-noise wave excitation.

In FAST 7.02, the way to define banded white-noise is to define your own wave spectrum through the UserWaveSpctrm() routine, which will require a recompile of FAST. Because the spectrum you want is white (constant in magnitude over a desired frequency band), the routine can be quite simple. In FAST v8, white noise is one of the default wave models, and so, is trivial to enable.

Best regards,