Floating Objects in DualSPHysics

edited March 2013 in Old versions
Dear,

first of all, as posted also at other places in this forum, congrats on this projects. Impressive seems appropriate.

I have a question on the implementation of floating objects in the released code.

Is the implementation in the released code the same as the one in the paper of Omidvar et al (2012) "SPH for 3D floating bodies using variable mass particle distribution", apart from the variable mass particle distribution to obtain a fine resolution at the floating body?

And, next to this paper, are there other validation tests?

Thanks for your efforts,

Sincerely,

Tomas

Comments

  • Hi Tomas

    The implementation is the same with SPHysics, but look like there is a bug in the code. The code currently does not calculate the weight of the floating objects right. I have had some communications and the developers are working on this problem to fix it. Also you might know that the code does not work with 2D floating problems.

    Cheers,
    Jalal.
  • edited April 2013
    A quick workaround I found was if you match the lattice of both the fluid and the boundary (currently using one for both) to be the same you appear to get the correct weight as the right equilibrium high due to bouyancy of obtained under particle refinement.

    If you follow the example of casefloating and use lattice bound="2" and fluid = "1" the object behaves like it has twice the weight of the fluid. (I think the fluid might be less dense by a factor of two than the object is heavy in this case useless I missed a scaling in the massfluid value somewhere)

    Mark.

  • Hi.

    I had the same problem with floating objects.
    If you place the floating object in the right position, you have to check the quantity of fluid particles whitch are created. Then you remove the floating object check the quantity again. The difference between the number of fluid particles is the value you have to multiply with the "massfluid value". So you get the "weight" you have to set as "massbody value" for the floating object.

    Sebastian
  • edited April 2013
    Hi Mark and Sebastian

    To get the xml file show the right weight in the dual-SPHysics input , I used the relation relativedensity= (floating mass/(No. of floating particles * mass of floating particles)). this is irrespective of massbound or lattice number. but when I solve the problem, the floating object behaves much lighter than it should be.

    Sebastian: My floating object is actually heavier than water and it sinks after dropping to the water. I have tried giving higher massbody values but I am lost with the ratio. It differs with lattice number and massbound. By "right position " do you mean with the right draft (sinkage)? As I work with external files and input vtk files for my floating object.

    Cheers,
    Jalal.
  • Hi Jalal

    The problem with a floating case is that the "floating-boundary particles" are not exactly on the boundary. So if you draw only a floatingbox with 1x1x1m you will see in paraview at timestep "0" (with lattice 1 or 2, and if you output with "partvtk" -onlytype:-all,+floating) that the box which was created by nodes is bigger than 1x1x1m.
    So if you want that the box swims 50% out of water the "massbody value" is not 500kg because the box displaced more than 500kg fluid.
    If you want that the box swims with 50% out of water, you have to start 2 cases to check how much fluid-particles are displaced by the box and to calculate the the mass you have to set.

    First place the box in one case with 50% out of water.
    In a second case you delete the box.
    Now you check the difference between the "fluid-particles".
    Because of this difference you are able to caldulate the mass of the fluid, which is displaced by the box (difference of fluid-particle * "massfluid")
    And so: If you want that the Box swims with 50% out of wather you have to set the calculated value as "massbody value".


    Sebastian
  • Thanks Sebastian.
  • New version 3.0 includes an improved formulation of Floating Bodies since the old version and the fortran versions includes some bugs that have been fixed now.
  • edited September 2013
    hi, thanks for the hard work, and the solver is very good.

    I understand that in version 3.0, there is some improvement on the floating object. I run the v3.0 beta version. But I still have some question here. What is the use of relativeweight="0.8"? In addition, in CaseFloating case, the mass of floating object if 2403 and inertia is 2681 both in V2.0 and V3.0. This cannot be correct. Because for cubic, the momentum of inertia is 1/12*m*(a^2+b^2), in this case, a=b=2, and this relation doesn't agree with the value above. Could you please brief a little bit about what change you made in version 3.0 regarding to floating object? Was the original bug you mentioned in gencase or dualphysics? Sorry about so many questions, but in general I think it all belongs to a big question, how is the mass,density of both floating and fluid particles involved in the computation. Thanks very much!!!
  • I also find that the interaction between two floating object particles is also contributed to the dv/dt of each in the code, which I doubt here. Could you please check it or maybe it has been corrected in the V3.0?
  • Dear all,

    I'm sorry for replying only now, I'm doing it as fast as i can.

    The floating bodies implementation in V2.0 has several problems:
    -Mass and inertia are not correctly computed;
    -Fluid-rigid interaction was deeply flawed at the level of the conservation equations;
    -Rigid-rigid (or boundary) interaction is not properly resolved. I might look good kinematically but it is good behavior by chance.

    In V3.0 beta, the equations were corrected, so now fluid-solid interaction should be more accurate. Problems with very low densities come up as a consequence of the delta-sph implementation, it will be corrected for the final version, but problems remain with the current boundary conditions. Solutions are being looked at but it might take a while to give you something 100% satisfactory.

    Regarding the mass and inertia problems, first you must understand that the problem comes from Gencase, and that it is more of an artifact than a bug. We do the mass, inertia, centroid, etc computations on the discretized system, i.e, after the raw geometry is converted into a collection of particles. Because of resolution, most times a 1m^3 cube in the .xml is not exactly the same volume after discretization. This is the problem, everything is coherent after this. It is remarkably difficult to generalize this and make it accurate at the same time. The solution however is simple, instead of specifying a relative density, make it an actual mass:. This forces the total mass to be correct and minimal errors in the inertia tensor should be present (because the discretized geometry is slightly distorted, at least by one dp in any direction).

    For last but not the least, this issue has been adressed in the manuals of sphysics and dualsphysics, rigid-rigid contacts are not treated and require validation. This is precisely the larger part of my thesis and I'm counting on releasing an .exe with a bet version of my current solution in a few months, after the release of v3.0.

    Best regards,
    Ricardo
  • Hi Ricardo,

    I also have found various problems in floating bodies implemented in DuapSPHysics V2.01.
    In particular, I'm trying to simulate the behavior of a floating wind turbine, but in DualSPHysics (v2.01) the movement of the floating wind turbine is not real, it seems that the mass and inertia are wrong.
    Have you found a way to solve this problem?
    Are you also treating the rigid-rigid contacts between 2 floating objects?

    Thank you,
    Elvira
  • Dear Elvira,

    as Ricardo has explained, those problems have been solved in the new version 3.0

    the contact between 2 floating objects is been solved using other approach:
    Canelas R, Crespo AJC , Domínguez JM, Ferreira RML. 2013. A generalized SPH-DEM discretization for the modelling of complex multiphasic free surface flows. 8th SPHERIC

    Regards

    Alex
Sign In or Register to comment.