Optimizing EDEM Simulation for Mud and Limestone Particles
Hello Community,
I am currently working on simulating mud flowing through a chute using EDEM software. My simulation involves two types of particles:
- Mud (Limestone + Water) with a particle size of d < 4 mm.
- Dry Limestone with a particle size of d < 4 mm.
The flow material consists of 20% Mud and 80% Dry Limestone. However, I am facing challenges with the simulation time, which is extremely high, making it difficult to proceed efficiently.
To simulate the cohesive behavior of the particles, I have applied a JKR coefficient of 15 to both particle types and utilized the JKR model to account for their physical interactions.
I would greatly appreciate any advice on how to reduce the simulation time without compromising the accuracy of the results. I have attached my simulation deck for reference, and I would be grateful if someone could take a look and provide insights or suggestions.
Thank you very much for your time and support!
Best regards,
Answers
-
Hi Pablo,
The attached file consists of only .dem, I need .dem, efd, .ess, .ptf, .dfg and 0.h5 (inside the folder) to open and review the file.
Please share all the above files mentioned above.
Thanks,
Prasad A
0 -
Thank you Prasad for your answer,
Please find attached the files mentioned above.
0 -
Hi Pablo,
I received the files, working in it will update soon.
Thanks,
Prasad A
0 -
Hi Pablo
Based on crude calculations, it seems like you have/will have over 90 million particles in your system, which is a lot of particles to get any reasonable runtimes, in general, for any EDEM simulation, unless you have multiple H100 GPUs that you are running EDEM on, which will still take more than a couple days for the simulation to complete.
Please consider scaling up your particles (as described in https://doi.org/10.1016/j.powtec.2015.05.051). A larger particle will reduce the total number of particles and make your simulation faster. Please remember to calibrate your particles again after scaling them up. You can find more information about calibration here: https://learn.altair.com/course/view.php?id=513
Also, please refer to this article which mentions one of the ways to reduce simulation run time:
Finally, you can also take a look at these two videos which shows how the factory can be optimized for speeding your simulation:Hope all of this helps.
Thanks,
Jerrin Job0 -
Hi Pablo,
in addition to Jerrin's comment, the total number of particles to be created as >45 Million, This model requires high GPU memory i.e 40GB or 80GB to solve. i.e either A100/H100/H200.
Low performance GPU card doesn't give you computational gain, Also to reduce the computational time, you can increase the particle size but do remember that re-calibration to be performed.
One correction what i suggest is use 3-DImensional factories instead of 2D- factories
The problem you setup looks like dynamic flow of particles i.e filling of particles through factory and at the same time discharge is opened instead of filling of particles by closing discharge and opening the discharge port to allow particles to flow due to gravity what i can suggest as a last options is solve for intermittent mass i.e first create the half of the mass of the total current mass and discharge and the other half you can create and discharge. This method I suggest only as a last option and you should make sure that how the physics of the system affect due to partial loading and discharging.
Thanks,
Prasad A
0
Altair Employee