Altair RISE

A program to recognize and reward our most engaged community members

Nominate Yourself Now!

Particle Size Distribution not activating - EDEM

Hi everyone!

I am setting the size distribution for two particles: DRI and Anthracite. The measured distributions are as follows (in ferret diameter):

Now, when I set this distribution in the EDEM GUI and run my simulation, the particles are not actually created according to the distribution. Rather I only see a single size for DRI, as wel as a (different) single size for Anthracite equal to their physical radii.

So, is there a reason why EDEM is not implementing my PSD? Have I interpreted a setting wrong or is there another window that I cant find where I should activate PSD?

My EDEM settings are as follows:

DRI:

Anthracite:
The lower cap is such that the min radius equals 2.775*0.721 ~ 2mm.
The upper cap is ~9mm.

Please dont worry about my contact radius settings. These are specified such that after the DRI shrinks 25% the contact radius is exactly its inital radius. Similar for anthracite but then with a shrinkage of 10%.

Also, yes I want my contact radius to also scale with the same distribution.

But again, my post-processor shows the following:

Does anyone have an idea what I am doing wrong in my settings and why EDEM is not generating my PSD?

Thanks in advance!

Kind regards,

Jan-Thijn Wijnker

Find more posts tagged with

EDEM

Troubleshooting

Accepted answers

Eric_Veikle

Hello Jan,

From lines 119 to 127 it is not clear to me how you are appropriatly updating the value stored in the custom property - particle initial radius. You are calculating R0 - at the first timestep after you call the →getDelta. I would have expected that you need to calculate R0 and then use the →getDelta.

To your question regarding scaling of the particle. You must take into account the prototype particle physical radius when scaling. In the example you offered if the particle's physical radius is 3 mm but through the PSD applied to the particle it is created in the simulation with a physical radius of 2 mm → a scale of 0.66667 is applied to the particle. If you then set the the scale of the particle to 1 through the API the physical radius will be updated to 3 mm in the simulation.

You might also consider using the scale value of the particles directly instead of using the volume to back calculate the physical radius of the particle. Just note the radius of the particle in edem at each timestep is the prototype particle physical radius multiplied by the particle scale.

I hope this helps!

Kind Regards,

Eric

All comments

Eric_Veikle

Hello Jan-Thijn,

Your setup of the particle size distribution (PSD) appears to be correct. As a first step, I recommend generating and accumulating a sufficiently large sample size. Then, use the histogram tool in Analyst to report the PSD based on particle diameter.

The screenshot you shared does suggest two distinct, uniform particle sizes, which would be consistent with a bimodal distribution. However, it's not enough to definitively identify the issue.

If you can share the results here after running the analysis, and they confirm a bimodal distribution, I'd be happy to take a closer look and help investigate the root cause.

Kind Regards,

Eric

JT_Wijnker_11

Hi Eric,

First of all, thank you for your quick response! Maybe some background info is in order. I am studying the melting behaviour of DRI, flux and anthracite as they fall into a slag bath within an electric furnace. To do so, my EDEM model contains an API infrastructure that handles conduction, convection, temperature update and consequently particle melting.

I can not simulate for long times as the first particles in the simulation will start to melt and therefore disrupt our particle size distribution histogram. However, I think the following result speaks for itself:

Now, because I wanted to show you this plot with a larger sample size, I decided to deactivate my Particle Body Force plugin that handles shrinkage and removal of the particles (mimicking melting).

Weirdly enough, after deactivating this plugin, the particle size distribution is suddenly implemented as seen below:

DRI Histogram (Without shrinkage and removal plugin):

Anthracite Histogram (Without shrinkage and removal plugin):

So, even though the sample size is still low, we can see that DRI starts to follow a normal distribution and it seems correctly. The Anthracite distribution does not seem correct with what I would suspect from my GUI settings. It should look more like:

I dont know why we get different results for anthracite as the values shown in my earlier message correspond with the above fitted curve. Maybe the filled in GUI values are not how it was intended by DEM?

Now, my main concern is that I dont understand why the removal of my particle body force plugin results in the PSD to be activated. And if I run the simulation with this plugin active, there suddenly is no particle size distribution and we again get our bimodal results.

I have very carefully checked my shrinkage and removal script and I am very confident in the fact that there is nothing inside it that should disrupt the particle size distribution.

In general lines, the shrinkage and removal script reads the radius of the particle in the first time-step and stores this as "inital radius" custom property. Then in later time steps, if the particle temperature is in the melt interval, the initial radius is used to scale the particle and shrink it gradually. Now, for numerical stability, there is a maximum cap that can be set on the shrinkage to prevent particles that are too small. There is nothing in the script that would give me any reason on why a particle size distribution would not be created when having this plugin active. I have build the logic to be compatible with PSD.

I have attached the Shrinkage and Removal custom plugin for you to see for yourself. The script contains some comments so it should be easy to understand.

ShrinkageAndRemoval Altair Discussions.txt

If you have any idea's on how it is possible that a particle body force plugin prevents a PSD from being formed, please let me now.

Kind regards,

Jan-Thijn Wijnker

Eric_Veikle

Hi Jan-Thijn,

Might be good to double check that your custom particle property "Initial radius" is a non-constant value per particle type. Can you check? I am suspicious that you are not updating the value stored in the custom property.
Let me know…

Eric

JT_Wijnker_11

Hi Eric,

Yes I update the inital radius but only in the first timestep of the simulation for each particle individually.

In my calculateforce() function there is the following loop:

The "Initial Radius" property is initalised as 0.0. So, at the start of the simulation if a particle is generated and its initial radius property has not been set correctly, this loop is activated.

I have checked in MSVS debug and this goes correctly. So as far as I can tell, each particle is correctly assigned its "inital radius" custom property based on its volume at the start of the simulation, and this only happens once.

I do not understand how this could influence the particle size distribution upon generation. Also, my script only starts shrinking (or any scaling for that matter) if particles are reach high enough temperatures, which only happens after multiple simulation seconds. So at the beginning, there is nothing that would alter or counter a particle size distribution upon creation. I have extensively tested and verified this in MSVS debug mode with single size particle simulations to make sure the logic is correct.

Also, in a single particle size simulation, the post processor correctly shows the inital radius value in the post processor dependent on the particle size I set for the particle.

Does this help you or am I not clear?

Kind regards,

Jan-Thijn Wijnker

JT_Wijnker_11

As I understand it, externalForce() is called individually for each particle, so each particle should store its own initial radius (R0) based on its unique volume at the start of the simulation.

Also, even if R0 were somehow set incorrectly (e.g., the actual radius is 6 mm but R0 is set to 7 mm), the script would still apply a scale of R0 / R0 = 1.0 at the first timestep (because particle temperature is low, this is verified to work accordingly). So the particle's size wouldn’t be affected, and the PSD would remain visually intact at that stage.

This is because for the very first timesteps, the targetradius always equals R0.

For a simulation with two particle types, for example DRI and Anthracite, with both a (different) single size particle. I have validated that inital radius is correctly populated for particle type.

Also, when I load in the plugin under Particle Body Force Interactions but don't select any particle type the distributions are generated correctly. If I select only DRI and set its settings for this plugin, then indeed DRI is not correctly generated but anthracite is. And vice versa. So the loading of the plugin itself does not seem to disturb anything. But setting it active for a type does apparantly.

Does that give you enough information? Based on this alone I still dont see how my script could influence the generation of the particle size distribution. Do you have any other suggestions for me to investigate to rule out what the isue is?

Thanks in advance!

Jan-Thijn Wijnker

JT_Wijnker_11

Good News:

In my script I previously had the following:

This line was run at each timestep, even if the scale would just be 1. Just to test I changed it to the following:

Meaning that as long as there is no need to change the particle size (i.e. scale remains 1) we do not actually run this line.

Somehow, I now do see a particle size distribution being generated even with my plugin active for both particle types.

So my hypothesis from this is that the setscale(particle.ID, scale=1) somehow counteracts the scaling that EDEM uses in the background for particle size distribution generation. Does that make sense to you in any way?

This leaves me with one final question:

When I call setScale(scale) later in the simulation to shrink my particles, is this scale applied relative to:

(a) the PSD-assigned size at creation. So lets say the physical radius in the GUI is 3mm but the particle is created with 2mm radius according to the PSD. Does setscale(1) then equal 2mm or 3mm? Because I think this was the source of our problem.
(b) the base physical radius set as template in the GUI. (i.e. 3mm)

You see, my assumption was that the scale is relative to the diameter a particle has when entering the simulation, even for particle size distributions. But it seems that scale() always is relative to the physical template radius. Can you confirm this?

Jan-Thijn Wijnker

Eric_Veikle

Hello Jan,

Kind Regards,

Eric