Fluids

Fluids

DEM Hertzian Collision iteration time

    • TE_Haf
      Subscriber

      Hi,

      I am testing a configuration where I have a free jet of solid particles and air flowing against an inclined wall. To reduce the size of the domain I am using the symmetry of the jet to simulate just a piece of the domain with roughly 5 Million cells.

      Particles are injected when the fluid flow converged and updated every fluid iteration. The time step of the particle is 0.001 with one iteration per fluid iteration.

      For the particles I am using the DEM Hertzian model, but as soon as the first particles hit the wall there is no more updates in the console. After 20h (8 CPUs) only 3 iterations were calculated. If I check the particle tracks I can see that the particles achieve the maximum allowed velocity after the collision, what makes no sense, since it is much higher than the velocity previous to the collision.

    • DrAmine
      Ansys Employee
      There are few examples on the customer portal. For DEM the most important step is to provide proper collision parameters and to use a particle time steps 1/10 or even smaller than the collision time.n
    • TE_Haf
      Subscriber
      Thank you,n I've checked some equations for the estimation of the collision time and it gives me values as low as 10^-8 s, which is really low. nWith this time step I increased the number or particle iterations per fluid iteration to 1000, so that it would not be too slow. Now I have the problem that every particle time step particles are injected with a mass much lower than a single particle. Unfortunately, now it is not writing any particle information to the .xml file. This is probably due to a too low mass or to small time step.nIs there the possibility of changing the particle time step just prior to a collision? For example, if the solver predicts that there will be a collision, automatically decrease the time step just to calculate this collision and then return to a faster time step while the particles are just translating?Could you be more specific where I can find an example of the DEM model beeing used on the customer portal. I couldn't find it with the search function using DEM, or Hertz as filter.nThank you for your help!n
    • DrAmine
      Ansys Employee
      Bit hard to achieve and requires a lot of coding if possible.nYou can look after Examples on the customer portal. There is also also another example using DEM. Moreover there is a sort of DEM Best Practice. Search words are DEM and just show Fluent hits.nnIf the collision time is so small then better to decrease the spring constant or use large particles. Do not forget that the collision time is based on parcel diameter /mass and not on particle diameter /mass.n
    • TE_Haf
      Subscriber
      Thank you.nI found the DPM best practice presentation, and it was really helpfull. Reading some other posts helped me to understand how I can use the parcel size to increase the collision time. Unfortunately, I could not get the Hertzian model to work properly. nI think the problem is with my configuration. The relative velocity between particles is very low, but the relative velocity between particles and the wall is high, so I probably need different spring constants for the wall-particle pair and for the particle-particle pair. Using the equation for the spring constant I obtained 2000 N/m for the particle-wall case and k~1 N/m for particle-particle collisions, if I consider that the velocity difference is slow. nUnfortunately, sometimes I still have particles flying with the highest velocity allowed after colliding with the wall.nIf I use a small particle time step(1e-7), but a large number of particle iterations (1e4) between fluid iterations, will the collision be solved differently than when I have a small particle time step with a single particle iteration per fluid iteration?nIs there a way, with or without UDF, to:nIdentify the exact moment when a collision occured?nDebug the collision process (during particle iterations)?nDefine if a collision happened with the wall or with another particle?nWhat causes the particles to accelerate to max. velocity after collision?n
    • Rob
      Ansys Employee
      Yes: use the particle summary file for that surface. Not sure what you meannYes, using the summary, but that won't pick up the particle-particle collision. You may need a UDF for that.The rapid acceleration is caused by the time step being big enough (and is therefore too big) that the particle overlaps with the collision partner. The solver then calculates the energy based on the overlap & therefore the speed. n
    • TE_Haf
      Subscriber
      Debug the collision process (during particle iterations)?nNot sure what you meannWhat I ment is, how can I verify if the resulting velocities after the collision are physical or unphysical? For example, im my previous figure the velocity after the collision was at least twice as large as the velocity of the particles previous to the collision. That looks unphysical to me. nHow could I check the kinetic energy/Momentum before and after every collision to make sure there was no increase in energy, or filter unphysical collisions?nAnother approach, of decreasing the time step to verify the convergence of the results is also not possible, since this affects the injection frequency.n
    • Rob
      Ansys Employee
      If the speed is higher than it can be then it's probably not correct, and this tends to be a result of the time step being too large relative to the collision time. n
Viewing 7 reply threads
  • You must be logged in to reply to this topic.