Confusion about IceDyn

[Jason.Jonkman]

Dear Tianhui,

I'm not familiar with the PSD script you are using, but there appears to be a lot of windowing, which I do not believe results in a very realistic PSD in your case. Can you try a simpler PSD script, such as my script that calculates the PSD directly via FFT without windowing (but with optional binning to smooth the PSD). This script, Jason_PSD.m can be found in the following forum post: viewtopic.php?f=4&t=2219&p=12891.

Best regards,

[Tianhui.Liu]

Dear Jonkman
First of all, I apologize for taking so long to reply. I was waiting for your reply before and then I realized it was on the second page.
In order to better illustrate my process of processing the results, I have put all the graphs obtained from the process here. And with some basic descriptions:
In the process of processing the data I selected all the results after 100, the total calculation time is 400 is, the time step is 0.005, each data has 6000 values
The picture on the left half is the result in the time domain, while the picture on the right half is the result of my own calculation using fft
The four images correspond to
1.iceforce
2.Tower-top / yaw bearing fore-aft (translational) acceleration (absolute)(From ElastoDyn)
3.acceleration at node 1 of member 3 (located at 0 m, i.e. MSL)(From SubDyn)
4.fore-aft shear force at mudline(From SubDyn)

1.png (821.84 KiB) Viewed 93 times

The graphics output by using your script look like this(On the right is the generated discrete point plot, and on the left is the generated curve)

Code: Select all

 [f1,Sf1]=Jason_PSD(iceforce,0.005,8192,1);
 [f1,Sf1]=Jason_PSD(TTA(:,1),0.005,8192,1);
 [f1,Sf1]=Jason_PSD(TBA(:,1),0.005,8192,1);
 [f1,Sf1]=Jason_PSD(TowerF,0.005,8192,1);

I'm not sure if I'm using it correctly, so please point out if there are errors.
The first parameter is of course the data, the second parameter is the time step I chose for my calculation, and the third I filled in is 8192 (when in normal fft a larger value than the amount of data is taken here, and it is often the nth power of 2. Each set of data has 6000 values so I filled in 8192 here)

3.png (1.33 MiB) Viewed 93 times

Regarding the previously mentioned WAFO function for calculating the power spectrum, I tried to adjust the smoothness setting and got a relatively normal PSD.

untitled.png (83.27 KiB) Viewed 84 times

Whether I calculate the fft myself or use WAFO's function to calculate the power spectral density, one thing I don't understand is why the excitation load is concentrated at low frequencies while the response (e.g. YawBrTAxp) is concentrated at high frequencies.
Whether I calculate the fft myself or use WAFO's function to calculate the power spectral density, one thing I really don't understand is why the excitation load is concentrated at low frequencies while the response (e.g. YawBrTAxp) is concentrated at high frequencies. The script you gave also shows this (although the scale in the frequency domain is slightly different)


Best regards
Tianhui

[Jason.Jonkman]

Dear Tianhui.Liu,

It doesn't appear that you are using the Jason_PSD.m script correctly. If I understand correctly, your time series goes from 100-400 s (300-s long), with a time step of 0.005 s (for a total of 60000 points). I also think it helps to bin the PSD for smoothing (without binning, df = 1/300 = 0.00333333 Hz; with BinLen = 10, this would coarsen the result to df = 0.0333333 Hz). So, I would expect you to use Jason_PSD.m as follows:

Code: Select all

[ f1, Sf1, f1_Bin, Sf1_Bin ] = Jason_PSD( iceforce, 1/300, 60000, 1, 10)

How do these plots look?

Best regards,

[Tianhui.Liu]

Dear Jonkman
Following your advice, these are what I got:

JASON1.png (1.34 MiB) Viewed 58 times

JASON2.png (1.31 MiB) Viewed 58 times

I have to say, this looks a bit confusing due to the large number of data points.

Best regards,
Tianhui

[Jason.Jonkman]

Dear Tianhui,

In your original PSD plots (from WAFO), I see dominant peaks at:
0 rad/s = 0 Hz
15 rad/s = 2.4 Hz
34 rad/s = 5.4 Hz

In the plots from Jason_PSD.m, I see dominant peaks at:
0 Hz
2 Hz
5 Hz
15 Hz

Overall, the agreement seems quite close (the 15 Hz peak is beyond the range plotted in the WAFO results). It may help to show the results from Jason_PSD.m on a linear y-axis to better compare to the results from WAFO. Again, I'm not sure why you see oscillations near the peaks in the WAFO results. The Jason_PSD.m results don't seem to show these oscillations, at least not with the log scale.

Best regards,

[Tianhui.Liu]

Dear Jonkman
The oscillations near the peaks in the WAFO results seem to be related to the smoothness set in the WAFO function.
I am rather confused as to why the peak of the spectrum of the excitation load (the first graph) and the peak of the spectrum of the response (the other three graphs) do not match.
I can't see any pattern from this.

Best regard
Tianhui

[Jason.Jonkman]

Dear Tianhui,

Looking only at the results from Jason_PSD.m, the main ice load happens at 0 Hz (the mean load), with energy decaying until it plateaus at about 10 Hz. However, I do see "bumps" of energy every 2.7 Hz, which also show up in the structural response. I'm not sure why these show up in the ice loading (again, I have little knowledge of IceDyn), but they do seem to be propagating to the response.

Best regards,