Strange multiphase behavior


I've just noticed a strange behavior when trying to simulate a 3D multiphase flow (2 phases, densities 7000 and 2500) :

Two liquids with different densities are separated by a wall. One removes the wall, and theoretically, one gets a final situation where the most dense liquid lies at the bottom, and the other one on the top.

But, when the two liquids get in contact, they litterally explode as if a strong repulsion force was in action.

I can send the *xml file if need be (how do I proceed ?)




  • Some illustrations of the phenomenon : in yellow the high density fluid, in red the low density fluid...

  • Which solver did you use?

    On an unrelated note: What tool did you use to post process?
  • I'm trying to use DualSPHysics4.0_LiquidGas_win64.exe
    The post processor is Paraview 5.6.0 64-bit

    BTW : in certain (rare !) circumstances, the simulation runs normally. But a small change makes it crazy again. For example, changing the limits of the computational domain, or changing the particle size to a smaller one... etc. Three new pictures, taken at three time steps following one another :

  • I'm personally also trying to use DualSPHysics4.0_LiquidGas_win64.exe for a multiphase simulation of 2 fluids. However, I have also had extremely lacklustre results so far. My best results were achieved using artificial viscosity (while I really want to use kinematic viscosity).

    Did you have a look at the density profiles? (Might notice weird fluctuations)

    I just can't exactly wrap my head around the fact that it sometimes works, and if a very small change is applied it explodes.
  • benben
    edited February 26
    Hi Cdc73 and Bladomas1,

    Using the multi-phase code is not as simple as the single phase code. The main reason is that the physics is different (before an SPH model is considered).

    In your case, you are trying to use the gas-liquid multi-phase formulation (designed for air-water, 1 kg/m3 to 1000 kg/m3) for 2 phases with densities of 7000 and 2500. These densities are typically for solids where the densities and physics are different. None of the multi-phase SPH codes are designed for these densities.

    This explains why suddenly the code will explode, because SPH is trying to respond to the physics being imposed upon it, but will always fail because the physics of gas-liquid interaction is different. The physics of your application is not implemented in the SPH formulation. When you choose your SPH model there are lots of elements to consider, for example how the stress (or pressure for fluids) is related to the densities or strain rates? What physical process controls the timestep? Often CFL is simply not appropriate. How do you define the speed of sound both in the physics and then numerically? ...

    You need to choose the model with the correct physics before running a simulation.

    I hope this helps,
  • Thanks Ben for taking time to answer.

    In the Liquid-Gas version of DualSHPysics, the material description let the user choose what kind of fluid is to be used... "0" means liquid, "1" means gas. Nothing in fact prevents a user to consider two liquids instead of liquid+gas.

    I thought the software would considere my two liquids in the same way as one.

    The fact is that sometimes, it works ! My feeling is that everything is ready to compute two non-miscible fluids, but a small bug somewhere...
  • HI Cdc73,

    I think some checks in the code to prevent users being able to generate physically incorrect cases would prevent this and also the "sometimes it works" (even when it shouldn't).

    Best regards,
Sign In or Register to comment.