Pierson Moskowitz spectrum

Hello everyone,

I was trying to generate time series from a spectrum. In particular, i am generating wave elevation in time domain from PM spectrum (Pierson Moskowitz). In order to verify that i generate the correct time series, i should reconstruct the spectrum from the time series i have generated.

Please give your opinion regarding the reconstructed spectrum in comparison with the initial PM spectrum.

Here a figure for the time series generated from PM:

wave_elevation

Here a comparison between the reconstructed spectrum and Pm spectrum:

comparison_between_PM_&_constructed_spectrum

I was thinking of a normalization problem.

Best Regards,

Riad

Dear @Riad.Elhamoud,

A few clarifying comments/questions:

  • Are you using the HydroDyn module of OpenFAST to generate the wave elevation time series from the PM spectra?
  • Are you randomizing both the amplitude and phase of each wave component about the target PM spectra (WaveNDAmp = True in HydroDyn) or are you just randomizing the phase of each wave component, using the amplitude from the target PM spectra (WaveNDAmp = False) in HydroDyn.
  • Are you resampling the data in any way or are the time step/length and frequency step/length consistent?

Best regards,

1 Like

Dear @Jason.Jonkman ,

  • No i am not using HydroDyn module to generate time series. In fact, i am using MATLAB.

  • i am randomizing the phase only. The phase is a uniformly distributed random variable varying between 0 and 2*pi

  • In fact, i used the following parameters:
    frequency step=1e-3 Hz;
    time step=1e-2 seconds;
    Number of samples = 63001
    duration of the sample is 630 sec
    sampling frequency is 100 Hz

  • I did not understand what you mean by the time step/length and frequency step/length consistent.

Best Regards,

Riad

Dear @Riad.Elhamoud,

It is hard to offer much help when you are not using NREL tools, but I would generally say that if you using FFT-based methods, that the time step/length (dt/tmax) and frequency step/length (df/fmax) are interrelated. That is:

df = 1/tmax
fmax = 1/(2*dt)

Best regards,

1 Like