Smaller Dp Results in Noise in the Force on the Seawall

Hi all!

Greetings. I am modeling a 2D-Wave-on-seawall type of problem.

The model setup can be seen in the following picture where I have an inlet-zone to generate 2nd order regular wave and a trapezoidal seawall seated on a slope. The water depth in front of the wall is around 5.4m.

Wave height is 3.6m and wave period is 3.6s.

The time-steping algorithm implemented is Symplectic while two choices of dp results in quite different result for the force on the seawall obtained by ComputeForce :

The result of dp=0.3 is quite smooth and is without the overshoots while dp =0.15 seems unstable and is with "noises".

Given the conditon, I expect that the wave does not break and as a result no breaking-wave force is exerted on the wall, and even if there exists impulsive force, it can only account for the "ups" but not the "downs" in the time history if I am not mistaken.

So I think the result may come from the numerical part. From a PDE point of view, for smaller dp, we need smaller dt but I think this is guaranteed because we use CFL number to calculate dt.

And I also check the free surface elevation at x =30, and the result is quite similar and smooth (does this mean that the time integration is stable ?). Probably it's due to my BC settings or other parameters but I am not sure. I am wondering how to explain/resolve this and I am sincerely thankful for all of your insights.


Best,

Barry

Comments

  • Very interested in this, hope some developers would like to chime in. I think I understand theoretically why it happens, since it could just be that the finer resolution captures better higher frequencies - would be nice to know how to filter it properly then

    Thanks for making the post Barry!

    Kind regards

  • @barry

    As a side note, if you wouldn't mind, I would love to learn how you made the measurement graphics (green lines) in your Paraview picture, looks really awesome!

    Kind regards

  • Dear @barry,

    if your are using DBC, it might happen to experience such a noise in the force results, however it seems a bit to high to me in your case. Are you using artificial viscosity and DDT? if so for the latter, which option (Fourtakas or Fourtakas full?).

    Regards

  • Hi @Asalih3d :

    I am tring to play around with the CFL number to have smaller dt, also, I use the "double precision" to store the data, the result is better (but still not optimal.)

    For the measurements, in Paraview you can use "Ruler" function to draw the measurement.

    Best

  • Hi @iarba27 ,

    Thanks for your reply.

    Yes, I am using DBC, and I am on my way towards mDBC to see if this is going to resolve the noises.

    And yes, I am using Artificial (0.01) + DDT (Fourtakas, 0.1). And perhaps Fourtakas full is a better choice to include fluid particles that interact with boundary?

    Thanks for your reply in advance.

  • Please let us know if mDBC resolves the issue or not - I have a very strong feeling that it will not, since I believe this to be more fundamentally a problem with the SPH formulation. I would love to be proven wrong. Please also show your mDBC configuration (i.e. location of particles, normals poiting the right way), since that will have a huge influence as well.

    And thanks for the hint about ruler function in Paraview!

    Kind regards

  • Yes, use Fourtakas Full, please

  • Remember that Fourtakas Full (DDT=3= should be used when several layers (covering a distance of 2h) are used in all the boundaries

    Regards

  • Hi @Asalih3d,

    Yes, it's true that mDBC isn't resolving the noise that much, whereas Fortakas full helps to smoothen the curve. But I really like the "gap" mDBC gets rid of. I just followed the XML_GUIDE_mDBC for the settings and I used the Method 2 for configuration since currently mt geometry is simple. But I think I will need to use method 1 since I am going to dealing with complex geometry imported as stl.

    Best

Sign In or Register to comment.