Still Standing Water Moving Unphysically

Hello!

I have a small 3d case in which a cylinder is submerged in a small pool of water, where for some reason the water particles at the cylinder surface start to accelerate extremely and generate a high velocity. See picture:


This is of course very unphysical, but I am not understanding why it happens. I have a rather fine dp = 0.01, which gives roughly half a million particles, which should be more than enough for such a case. There is no initial velocity for the water what so ever, the only force existing in the domain is the gravity set to -9.81.

Has anyone simulated a similar case with no issues, if so could you share your case?

Kind regards

Comments

  • Would anyone be able to explain what numerically is happening since this kind of motion of particles are happening? I have found that the particles are even moving vertically up and not down, so it seems even more wrong.

    I have kept playing around with parameters like the speed of sound etc., but whenever I place an object fixed in a body of fluid an acceleration happens around the object.

    I hope to hear comments from some who have tried something similar and perhaps even found a solution.

    Kind regards

  • It seems like gravity is the issue, but that seems weird. Is it really impossible to simulate a still standing pot of water with a cylinder in it?

  • Hi,

    I have witnessed the same kind of behavior in the past. For me, it was with still or slow moving water and around floatings (no cylinders).

    Like you I have no idea why this is happening. My intuition tells me it is related to the dynamic boundary conditions, but it is just my intuition...

    Kind regards

  • @TPouzol thanks for chiming in! I was losing my sanity slowly, thought I was the only one who had this issue.. :-)

    I have been testing a lot of different combinations, without a final resolution, so I am a bit lost on what could be the solution, but I think the main culprits are the boundary conditions as you say, until hopefully the next version will have a way to strictly enforce no-slip condition in mDBC.

    I also notice that the initial pressure shock wave (which I see in almost every SPH simulation I make) produces unphysical horizontal velocity motion, while the gravity leads to interchanging of particle positions vertically, when they should just stay still in layers for 3D simulations.

    For some reason when including a fixed object into the water body it leads to a rise of force in vertical direction, which then further accelerates particle motion (IF GRAVITY IS ENABLED).

    Kind regards

  • Some questions:

    • is it the object fixed?
    • have you created the particles in such order that you fill the water volume as final step?

    Regards

  • Hi Alex!

    Yes, the object is fixed. I will check whether or not I have made water particles last tomorrow, perhaps that might explain things.

    I have made sure that the fixed cylinder does not penetrate the sea bed and also does not have a layer of fluid particles between it and the sea bed as well.

    Kind regards

  • edited February 2021

    @Alex for the case shown in the original picture I confirm that the cylinder is fixed and that the fluid particles are made last. When doing this I got forces of approximately zero in the horizontal plane, but about 10 Newton in z-direction, which I could not explain.

    It seems like this might be of importance even if no fluid particles are present inside the boundary. I have checked both with DBC and mDBC, and in general the issue has been bigger with mDBC, perhaps due to the freedraw mode, some fluid particles can get trapped inside.

    If you have any more tips please do let me know.

    Kind regards

  • Please share with us (dualsphysics@gmail.com) your XML and we will take a look when possible.

    Regards

  • received

  • @Asalih3d Late to this, sorry Can you share some additional information about these settings

    1. time stepping algorithm
    2. boundary conditions
    3. density diffusion formulation
    4. viscosity formulation

    that you have used in simulations that failed? Thanks. Perhaps there are some hints to get out of there.

    1. Symplectic
    2. mDBC mostly, also tried DBC
    3. Fourtakas 3 (Full), with default value of 0.1
    4. I tested both artificial viscosity and SPS-LES, not a huge difference between the two

    Kind regards

  • edited February 2021

    @Asalih3d I would suggest to try the density formulation of Molteni and Colagrossi first

    I think the DDT á la Fourtakas considers the density out of the total pressure, while that à la Molteni and Colagrossi considers the density of the dynamic pressure: the difference between the two pressures is indeed the hydrostatic pressure, which is proportional to gravity. Since you have seen that turning off the gravity helps, then it makes sense to try this out.

    Try this with both mDBC and DBC: it would be interesting to have a checkpoint after a comparison these two.

    A parallel avenue is to keep the DDT by Fourtakas, but double the h/dp you used: it will take longer because there are more terms in the kernel summation, but could stabilize the results.

    Intrigued by how it goes. Hope this helps clearing the mist.

  • If you plot the initial condition using parallel projection and top view you can see that the cylinder has not a symmetric condition in X and Y directions... and this will lead to initial motions so that to those initial values of forces.

    You have to consider the following:

    1) dp=0.04 may not be a good number to be used, as for example 0.01, 0.02 or 0.05 will be, I meant that perhaps the number of particles created for one dimension divided by 0.04 can vary +/- 1 particle so that we have those strange shapes, not symmetrical

    2) the other issue is the use of <setfrdrawmode auto="true" /> with low resolution...


    Those two issues are the important ones since they are key to defining your initial conditions, which is not symmetric now in X and Y as you can see in the figure, ok?


    Regards

  • edited March 2021

    @Asalih3d Seconding Alex' remark, you could take a step back and draw a prism (with a square base, aligned with the lattice) instead of a round cylinder, and remove the influence of the initial design in a simple fashion. So you can isolate whether the issue occurs either in full or with a different intensity or not at all. Hope this helps further.

  • Thank you @Alex and @sph_tudelft_nl

    I will look into this the next few weeks and return with my findings

    Kind regards

Sign In or Register to comment.