Unwanted oscillations in output

Hello! In this section we'll talk about the post processing module and the data visualization.
Post Reply
Carlotta
Posts: 6
Joined: Fri Apr 01, 2022 2:40 pm

Unwanted oscillations in output

Post by Carlotta » Sun Aug 07, 2022 3:32 pm

Dear team STKO,
I built a model of a soil domain formed by a soil deposit on a bedrock layer (the geometry was built with Python API and then I did the merge operation).
As input, I used a Ricker Wavelet with dominant frequency 0.6 Hz and PGA 0.2 g: I considered an acceleration history at the free-surface (outcrop motion), I used deconvolution process to obtain bedrock motion and I applied the velocity-time history obtained from a deconvolution.

I ran the dynamic analysis and I had some problems with the results:
when I extract the results at the bedrock points (selection sets "bed_f_ .." ) I do not find the Ricker Wavelet used as input but there are unwanted oscillations in the firsts time steps, only at the midpoint of the model there are no oscillations; the same happens at the interface points and at the surface points.

I used Newmark integrator and I had done many tests changing gamma and beta values: if I increase these values, it seems that the fluctuations decrease but they are still present.

I don't understand whether it is a modelling problem or a problem in choosing AnalysesCommand's attributes.

The velocity-time history that I applied with the AbsorbingElement in the X direction is applied at the base of the model only or at the vertical elements also?

Image

Below I attached the files of the whole model and an excel file that I used to process the results.
Thank you in advance.
Attachments
2D_DY_B_06_02_L.zip
(1.23 MiB) Downloaded 9 times

STKO Team
Posts: 1346
Joined: Tue Oct 29, 2019 8:45 am

Re: Unwanted oscillations in output

Post by STKO Team » Mon Aug 22, 2022 12:28 pm

The problem was related to the initialStateAnalysis:

1) You forgot to assign the InitialState material wrappers -> In this way the model will "jump" back to its initial position without keeping the initial stress, so it will generate oscillations

2) When you use absorbing boundaries after you reset the displacement to 0 (initialStateAnalaysis OFF), you need an extra "dummy" analysis (either static or transient) just because the absorbing boundaries are not awere of the initialstate material wrapper, so you need an extra step where the displacement are reset to 0
FINe.png
FINe.png (139.77 KiB) Viewed 253 times
Here's the modified file
2D_DY_B_06_02_L.zip
(277.32 KiB) Downloaded 6 times

Carlotta
Posts: 6
Joined: Fri Apr 01, 2022 2:40 pm

Re: Unwanted oscillations in output

Post by Carlotta » Thu Sep 08, 2022 5:23 pm

Thank you for the reply!
I have one more question if you don't mind:
the results we get are the same as if there was only one layer of bedrock (1); I would have expected something different like what I get in the model I have attached below (2), where I changed the values of:
- Soil Elastic Modulus
- Bedrock Elastic Modulus
- Shear Modulus in Absorbing Boundary
(but these are not the values I should use).
Image

It seems that the first model does not feel the difference between the two layers.

Also, in the second model, I do not find the same ricker wavelet used as input in the base node (2_node 61).

How can I fix these results?
Thanks for your support.
Attachments
prova5_forum.zip
(863.06 KiB) Downloaded 5 times

STKO Team
Posts: 1346
Joined: Tue Oct 29, 2019 8:45 am

Re: Unwanted oscillations in output

Post by STKO Team » Fri Sep 09, 2022 11:12 am

I don't fully understand your question.
However, in the file you sent me, you don't get the expected input at the base just because the bedrock domain is too small.

Given the material properties and the input frequency, the wavelength is about 6 km! you have to make the soil domain larger otherwise you will not be able to see the full wave entering the system:
wlen.png
wlen.png (49.14 KiB) Viewed 213 times
Here's the file:
2D_DY_B_06_02_L.zip
(324.04 KiB) Downloaded 6 times

Carlotta
Posts: 6
Joined: Fri Apr 01, 2022 2:40 pm

Re: Unwanted oscillations in output

Post by Carlotta » Mon Sep 12, 2022 5:22 pm

Thank you for your prompt reply!

So I could increase the depth of the bedrock or use a ricker wavelet with a shorter wavelength as input, is that right?
In STKO's pre processor is there the possibility to know what the input wavelength is, given the material properties and the input frequency? Or is it only visible in the post processor?

Thank you for all your assistance.

STKO Team
Posts: 1346
Joined: Tue Oct 29, 2019 8:45 am

Re: Unwanted oscillations in output

Post by STKO Team » Wed Sep 14, 2022 7:30 am

So I could increase the depth of the bedrock or use a ricker wavelet with a shorter wavelength as input, is that right?
Yes
In STKO's pre processor is there the possibility to know what the input wavelength is, given the material properties and the input frequency? Or is it only visible in the post processor?
It's just
lambda = Vs/f(shear wave velocity over frequency of the input motion)

Then your maximum mesh size should be approximately lambda/8, in order to properly represent that displacement profile in your mesh...

Carlotta
Posts: 6
Joined: Fri Apr 01, 2022 2:40 pm

Re: Unwanted oscillations in output

Post by Carlotta » Wed Sep 14, 2022 5:33 pm

Given the material properties and the input frequency, the wavelength is about 6 km!
If the shear wave velocity of the bedrock is 800 m/s and the frequency of the input motion is 0,6 Hz (I have to use the peak frequency, don't I?) lambda should be about 1,3 km , isn't it?

Thanks for your patience.

STKO Team
Posts: 1346
Joined: Tue Oct 29, 2019 8:45 am

Re: Unwanted oscillations in output

Post by STKO Team » Thu Sep 15, 2022 8:02 am

If the shear wave velocity of the bedrock is 800 m/s and the frequency of the input motion is 0,6 Hz (I have to use the peak frequency, don't I?) lambda should be about 1,3 km , isn't it?
imagine your input has typical frequencies in the range fmin-fmax
I would use fmax to compute the maximum allowable mesh size:
mesh_size_max = Vs/fmax/8 (in fact it gives the minimum value for the mesh size)

and fmin for computing the largest wavelength that can enter in your domain. However this is aspect is less important. I told you this just because you are doing a benchmark with a wavelength that has 1 main frequency. But with a real record it might be difficult

Post Reply