Effect of number of parcels and particles/parcel in DPM on results

[scabo]

Hi

I am simulating unsteady DPM simulations and I am trying to set the correct number of parcels and particles/parcels. How should we determine the number of parcels and particles. Do I need to do a sensitivity around that? And is the volume fraction of the discrete phase in the pipe domain is calculated based on number of parcels or particles? I need to keep the solid vol fraction around 1e-6.

[KM]

A sensitivity analysis on the number of parcels and particles per parcel is recommended.
The volume fraction depends on the total number of particles, not parcels, since parcels are just computational representations.

This solution could help to clarify how particles per parcel are determined.
https://customer.ansys.com/s/article/FAQ-2054353

[scabo]

 

Hi- i have attached the picture of summary injection. There the ‘total number of particles’ is per injection point right? Time step=0.0005 and time=0.5 s. Also to have a very low vol. fraction(1e-6) in a long pipe, i need to keep the number of particles/parcel <1 with 100 parcels/time step. Is it technically accurate to keep < 1 particle per parcel because normally number of particles>no of parcels.

 

[scabo]

 

And to get statistical convergence for particles I have plotted vertex avg of dpm-conc(Y axis) at a point with time (X axis). Does this look fine? And Do i need to increase the number of parcels or particles for convergence?

 

[scabo]

The convergence of vertex avg dpm velocity(Y axis) vs time(X axis) shows oscillations about a mean. Is this toleracne acceptable?

[KM]

The specified volume fraction and particle mass determine the number of physical particles. The number of parcels is only a computational parameter. It does not make sense to inject more parcels than physical particles.

These plots are difficult to judge. To properly study the sensitivity, it is better to compare converged solutions for the output quantities of interest.

[scabo]

Hi, Through these plots I am trying to see the convergence behaviour of dpm-velocity mag and dpm-conc at a point in domaimn(vertex avg). Do these plots look fine or will they convergence in a horizontal line like fluid systems? The X axis is time and Y axis is dpm-values as mentioned above. thanks

[KM]

This is not standard practice to judge convergence based on thess kind of plots.
It is much more informative to track integrated physical quantities (e.g. drag) and ensure that they stabilize over several iterations within each time step.

[scabo]

You mean drag force on particles? I want to see the statistical convergence of particle phase in dpm.