Fluids

Fluids

Rule of thumb – DPM #of parcels for my system?

Tagged: 

    • ai0013
      Subscriber

      Hello everyone,


      I was wondering if anyone here knows about a "rule of thumb" to determine an appropriate parcel count to obtain statistically significant results in my DPM simulation ?


      For example, my system consists of:

      Mass to inject: 5g

      rho_p = 1400 kg/m3

      d_p = [10-100] micron


      How many parcels would be advisable? I want to test the effect of the extreme values d_p = 10 and 100 micron.. If I keep a fixed number of parcel in 1 Million, then:

      Lower bound (10 micron) - 6.8e+9 physical particles -> Particle-to-Parcel ratio ~ 6.8e+3

      Upper bound (100 micron) - 6.8e+6 physical particles -> Particle-to-Parcel ratio ~ 6.8


      Is there any recommended particle-to-parcel ratio? I would appreciate any reference if possible

    • Rob
      Ansys Employee
      I tend to leave Parcel settings as default and then increase the number of stochastic tries to increase the number of parcels in the domain. I also use unique injections of fixed size over RR distributions so I know how many of each size are in the system.
    • ai0013
      Subscriber
      than you for your kind answer.
      My question is more related to know if theres any rule or formula to estimate a statistically-significant particle-to-parcel ratio?. As you can see, for a fixed dpm mass a decrease of 1order of magnitude in d_p, involves a factor x1000 in the particle-to-parcel ratio. Is there any threshold of this ratio for which the #parcels become unsufficient to accurately represent the physical particles in my system?

    • Rob
      Ansys Employee
      Not really, as you also have a parcel mass to cell (facet) size for momentum and erosion etc. Very simply you need enough parcels such that the particle trajectories can be seen and are sensible, ie modelling a spray cone with 2 streams isn't any good, 20 is probably OK and 100 is likely to be excessive but that will also depend on the size of the cone etc.
Viewing 3 reply threads
  • You must be logged in to reply to this topic.