Is there a set of “default” (read: suggested starting point) PIC parameter settings for setting up a new case. I have something like this (below) but I’m not sure if this is a good default or if I’ve adjusted these out of whack for some problem years ago.
What you posted is a great starting point for PIC settings and should work well for most simulations.
pic.close_pack should match whatever corresponds to the solids you are modeling.
pic.parcels_per_cell_at_pack can be thought of in terms of resolution with 24 being an approximate lower bound. This input controls how much effective volume each parcel represents. If you use a small value (e.g., 4), each parcel represents a lot of solids volume. Therefore, small changes in parcel positions will cause large swings in the solids volume fraction, and in turn, the solids stress.
I hardly ever use different values for beta an pressure_coefficient. The same can be said about the damping_factor settings.
As for the rest – I do not recommend changing them unless you have a good understanding of how they affect the PIC update algorithm. Arbitrarily setting initial_step_type, advance_vel_p, or velocity_reference_frame to different values most likely will not end well.
Comparing with a “regular” MFiX PIC setup, the default values seem to differ a little bit, added in bold below. Should these be used instead, or are these just different across the two editions? I’m setting up a Geldart group A PIC case, not much different from the NETL paper, but I get quite fluctuating pressure reading, and far less bubbles than found in say TFM. As the NETL paper keeps P=100, gamma=3 and statistical weight = 1000, I’m trying to see which other parameters that should be tweaked. Could it be damping, or something in the advanced section?
I would start with the values I provided. parcels_per_cell_at_pack is not statistical weight, it’s how many parcels it takes to fill a fluid mesh cell up to close_pack. You have to back out what the statwt is yourself. I doubt the wall damping factors have much of an impact at all. TBH I don’t understand the reference frame nonsense, so I never adjust that value.
Since you have a TFM curve, I’m assuming that’s from MFiX classic. Do you also have a PIC model with MFiX classic? Keep in mind, this is not just a new edition; it’s an entirely different code.
Thanks. Yes, forgot to mention, all of this is MFiX classic. Been looking for advice on the PIC setup, and thought this post could be of help. So my bad, I didn’t know the difference was that big between the two editions. (Although still a little surprised that the default PIC parameters are not the same.)
PIC can have particle size distributions. Without looking at the code, I assume it loops over each discretized size of your PSD and fills cells, so each fluid cell will have one size of parcel and the statwt will depend on size.
@wfullmer is correct that the statistical weight is determined by the parcel size.
A helpful way to think about PIC parcels in MFIX‑Exa is that they are material points representing a certain amount of solids material. I strongly recommend not thinking of them as particles or collections of particles. In general, we want each parcel to represent the same solids volume, so that when parcels move, they contribute consistently to the solids pressure term. This is why MFIX‑Exa uses the number of particles at close pack as an input—this parameter is intentionally decoupled from particle diameter.
As an additional note, each time a PIC parcel (or a DEM particle, for that matter) is created, its diameter is sampled from the specified particle size distribution.
