Breakage PSD Comparison with Experimental D90 and Agglomeration

Question

Hello everyone,

I am currently performing a simple breakage test in EDEM and trying to compare the simulation results with experimental PSD data. I would appreciate some guidance regarding the most appropriate way to process and compare the particle size distributions.

In the experiments, I only have D10, D50, and D90 values (volume-based), although the exact measurement methodology is not fully described in the paper.

In EDEM, when using the breakage model, I can obtain PSD information in two different ways:

1. Using the standard particle size distribution output, where I can plot/export volume-based PSD or extract CSV data and calculate D10/D50/D90 myself using Python.

2. Using the “Breakage Particle Size Distribution” output file generated by the breakage model, which gives size-cell and mass information. I can again convert the mass distribution to a volume-based PSD and calculate D10/D50/D90 from that.

The particles are cohesive, and I am using the Hertz-Mindlin with JKR cohesion model.

My main question is:

Which of these two PSD outputs is the correct/recommended one for comparison with experimental D10/D50/D90 values?

There is also another issue that confuses me:

My maximum primary particle size is around 500 µm. However, in some experimental conditions, the reported D90 is around 700 µm. This suggests that agglomeration is occurring in the experiments, meaning that some measured structures are larger than the original particles.

However, when I calculate the PSD directly from the EDEM particle data (volume or number based), it seems that agglomerates are not being considered as larger entities. The PSD only reflects the sizes of the individual particles/fragments themselves.

Because of this, I feel that I may never be able to properly match the experimental D90 values if the experiment inherently includes agglomeration effects while the EDEM PSD does not.

Could anyone please advise:

Which PSD output should be used for proper comparison?

How should agglomeration be treated in EDEM for PSD comparison?

Is there a recommended methodology to compare DEM breakage simulations with experimental volume-based PSD measurements when cohesive/agglomerating particles are involved?

Any suggestions or best practices would be greatly appreciated. Thank you very much.

Fatemeh Hosseini_22198 · Answer

Hi Stephen,

I have one more question regarding this.

Does this mean that when the breakage model is activated in EDEM, it is not really possible to properly study agglomeration behavior at the same time?

Since the “fines” are not physically generated and the PSD is only estimated, I am wondering how agglomerates can be correctly identified or tracked, especially when cohesion is involved.

Thank you very much for your help.

Kind regards,

Fatemeh

Fatemeh Hosseini_22198 · Answer

Hi Stephen,

I actually wrote a code to identify the agglomerates and calculate their sizes automatically. However, the issue is that the code is coupled with the actual EDEM particles/data existing in the simulation, not with the “virtual” fines estimated in the Breakage PSD output.

Because of this, the agglomerate analysis does not give fully correct results, since the fines are not physically generated in the simulation. On the other hand, the exported Breakage PSD file only provides sieve size and retained mass data, so I cannot directly identify or reconstruct the agglomerates from that information.

I was also wondering about another possibility. At the moment, my initial PSD is around 400–500 µm, and in the Tavares model I currently use a dmin of 15 µm. If I decrease dmin significantly, for example to 1 µm, would these smaller fragments actually be generated in the simulation?

If so, then I could potentially use my agglomerate-analysis code directly on the simulated particles and calculate d10/d50/d90 based on the agglomeration behaviour rather than relying only on the estimated Breakage PSD output.

Do you have any suggestions on how this could be handled more appropriately?

Kind regards,

Fatemeh

Fatemeh Hosseini_22198 · Answer

Hi Stephen,

Thank you for the clarification.

The material in the article is paracetamol crystals, and the PSD was measured using a Malvern Mastersizer 3000 with the Aero S disperser. The particles undergo high-velocity impact at the elbow of the disperser, which causes breakage/agglomerate dispersion, and the resulting PSD is measured immediately by laser diffraction.

Regarding the fines, that makes sense. So for comparison with the article, I should use the exported “Breakage Particle Size Distribution” rather than the generated particles only.

Kind regards,

Fatemeh