Particles Out of Simulation - Density Diffusion Option 3 (Fourtakas 2020)


Summary: Used New Density Diffusion Option 3, around 4% particles out of the simulation, want to know whether pressure -> force evaluations are reliable from such a simulation?

Hello,

I am running a simulation where a body is moving inside a fluid domain (with Chrono hinge). As I was working with different options, I set Density Diffusion Delta to Option 3 - (where it is applied to both fluid-fluid and fluid-body interaction)

I got a warning

*** WARNING: It is advised that selected DDT: 'Fourtakas et al 2019 (full)' is used with several boundary layers of particles when h/dp>1.5 (2h <= layers*Dp)

I did not think that much of it as the simulation ran successfully and I got my force output, with a little high frequency noise.

However, I was looking at the run.out file and noticed that a number of particles are out of the system

In the end, a little over 4% of total particles were out.

I am wondering about why these particles are out of the system and how much it will influence the pressure, and thereby force outputs?

I have attached the run.out file

Finally, I have fixed the issue using a multilayered boundary as suggested by the warning. Since each simulation takes a while to run, I want to know

if I have to re run all of them?


@Alex

@Asalih3d have you looked at this option before?



Comments

  • Hello,

    I have worked with moving objects multiple times trying different density diffusion terms, my observation is that shifting correction directly affects the number of particles that leave the domain.

    I can see that you have the shifting correction disabled, maybe you could try different shifting options.

    Secondly, as you have pointed out when working with moving objects, the results are pretty sensitive to boundary definition.

    If it is not a super high resolution simulation (in your case there are 800k particles, it would be painful to create .vtk files), you actually can see those particles leaving the domain in Paraview. However, my suggestion would be that if your results are close enough to whatever reference value you are using, do not really worry about those particles.

    In my tests, I had about 7% of the particles leaving the domain, and the effect in terms of mean drag and lift coefficients was less than 5% which to me was not really significant.

  • Hi!

    I believe I mostly use option 2 or 3 for Fourtakas DDT. Option 3 is the most difficult one, since it applies to all fluid particles if I remember correctly, while option 2 is only at the boundaries (i.e. interaction with solid objects), which is the most important location to me.

    To do good DualSPHysics SPH, you pretty much have to use density diffusion in combination with mDBC. So I would keep striving in that direction.

    I do not know your case, so I cannot tell you if 4% out is a problem - to me it would be.

    Kind regards

  • Just to clarify...

    Option 2 of DDT Fourtakas applies density diffusion term to fluid particles that do NOT interact with boundaries...

    while option 3 applies to all fluid particles, including the ones interacting with boundary particles...this is why using a number of layers to complete 2h range of interaction is advised with option 3

    On the other hand, the use of mDBC already needs to use several layers so that mDBC combines perfectly with DDT option 3

    Regards

Sign In or Register to comment.