Particle Scale Up Summary Blogs

Carles

Computational time has always been one of the most important aspects of DEM Modelling. Initially, it could be assumed that having an accurate and realistic material model through material calibration is the main target. If you want to learn about material calibration, read this post:

Discrete Element Method Calibration with EDEM

Stephen Cole

Mar 22, 2023

Discrete Element Method Calibration with EDEM

Calibration or the use of calibrated models is an important component of any Altair® EDEM™ simulation.  This blog summarises the most common discrete element method calibration approach and outlines the resources available.

Why calibrate?

Bulk materials come in all shapes and forms, a material such as coal from one area of the world may behave differently from another part due to local conditions such as moisture content and humidity.  In addition the processing of this material may be different resulting in unique size distribution and amounts of fines and therefore different bulk behaviour.  For powders one companies processing methodology and additives may be different to another, resulting in unique material bulk behaviour.  For accurate numerical modelling of these materials calibration is required against a known experimental measurement.

For more in-depth calibration knowledge please join our regular training sessions or review one of our webinars:

• Register for EDEM training
• Review calibration webinar
• Simple and Efficient Modelling of Powders with the EDEM Powders Database

For more background into EDEM we have a series of e-learning courses:

• New to EDEM ? – Check the EDEM eLearning courses

Calibration methodology

The typical calibration approach is to replicate the conditions of the process we wish to simulate in a small scale lab experiment, then through parameter variation and optimisation obtain our calibrated material model. Key steps are:

• Understand and classify the material
• Understand the process
• Characterize the material with an experimental test
• Reproduce the experimental test and analysis method in EDEM
• Perform parameter optimization to replicate experimental result

Understand and classify the material

The best way to classify a material is to measure its size and size distribution as well as its shape and shape distribution.  What do we know about the material already, is this a cohesive or free flowing powder, a series of irregular shaped rocks or a semi-rigid solid block? 

https://us.v-cdn.net/6038102/uploads/3AVGU1EKYCZJ/image.png

Coarse particulate solids can often be modelled at or close to the physical particle size with an approximation of the shape.

• Which EDEM particle shape to use ?
• Introducing Sphero-cylinders

https://youtu.be/dtyvbutcIm0?feature=shared

Fine particulates typically require a different approach as often it is not practical to model the required number of particles if the requirement is to model the complete system, the material is modelled using larger representative meso-scopic particles.

https://us.v-cdn.net/6038102/uploads/73IZ6GJUXRFQ/image.png

Understand the process

We need to understand the process to determine if we can model this system directly or reduce the system down to a manageable level.  How many particles are in the system and how long do we need to simulate for?

• EDEM Simulation Run Time
• EDEM GPU
• EDEM Benchmarks

What is happening in the system? Are the particles the dominant physics in the system or do we need to consider adding additional components such as :

• Simulating Fluidized Beds Using CFD-DEM Coupling
• Bulk Material and Multi-body Dynamics Simulation for Heavy Equipment

We also need to know the stress state and flow regime in the system of interest.  Is this a semi-static hopper of material under pressure or a material or a dynamic free surface flow in a conveyor transfer system.  The stress state is critical in understanding which experimental test to choose.

Characterize the material with an experimental test

Once we understand the process and the material the appropriate test can be chosen. The larger particle types often are in a more free flowing system and the system can be captured with angle of repose and inclined plane tests.  Powders maybe dynamic and free flowing or semi-static and under high or low stress in which case the appropriate test needs to be selected.

Calibration tests for Powders

https://us.v-cdn.net/6038102/uploads/SMIGAAAM0OFV/image.png

Calibration tests for Soils

https://us.v-cdn.net/6038102/uploads/MYYCZOBWH47J/image.png

Understanding the material and process also allows us to choose from one of the pre-defined already calibrated material models in EDEM:

• Material Models Libraries in EDEM
• Simulating soft soils in EDEM with the Soils Starter Pack
• Simulating powders in EDEM with the Powders Starter Pack (Legacy powders pack)
• Simple and Efficient Modelling of Powders with the EDEM Powders Database (Introduced in EDEM 2022.0)

Reproduce the experimental test and analysis method in EDEM

Finally we are ready to replicate the test, you can choose one of the tests mentioned above or create a custom test for the application. 

EDEM has a series of pre-defined tests with automated analysis. For an overview of the tests and kits available see:

• EDEM Calibration Kit
• Modeling a shear cell test with EDEM
• Material model calibration using Static Angle of Repose test with EDEM

The analysis method needs to match the experimental output, this is possible with the default analysis methods in EDEM or can be automated using EDEMpy.  The calibration kit includes pre-defined analysis scripts, for more information on EDEMpy see the e-learning:

• Introduction to EDEMpy eLearning

Perform parameter optimization to replicate experimental result

The common approach is to run simulations of the test multiple times, varying the key interaction parameters each time to match the response of the simulation model to the physical measurement.

https://us.v-cdn.net/6038102/uploads/OXO43PV7YJZN/image.png

HyperStudy and AI are powerful tools for parameterising and optimising EDEM models:

• HyperStudy Connector
• Combining EDEM with romAI-An efficient approach to process, product, and workflow optimization

There are a number of tools to help you get started with this, for example:

• EDEM Tutorials
• How to Calibrate Material Model with EDEM and HyperStudy (YouTube)
• EDEM Batch
• EDEM Physics Models e-learning
• AI-powered Solution to Speed-up System Design and Optimization

If you are looking for further information on EDEM or other Altair products we have plenty more on Altair Community:

Altair Community EDEM

 

---

View Post

However, even though calibration is always the right approach, it serves no purpose if the model cannot be calculated in a reasonable time, which is a subjective concept. Moreover, it does not matter who you ask nor the application they are modelling, faster simulation times are always desired.

There are many factors affecting computational time: from modelling aspects such as the chosen particle size to hardware attributes like the GPU used for calculation. For the latter, the user has often little to non flexibility on which card to select or its setup. On the contrary, user input is key to define particle size and it can come down to experience or feasibility to define a 1 to 1 relationship between DEM particles and real particles, which is very often impossible. However, 1 to 1 size scale is closer every day:

https://community.altair.com/discussion/32926/breaking-the-billion-particle-simulation-barrier-with-altair-edem

In this blog post I would like to summarize, list and share all the scale up processes and techniques recorded within the EDEM community so that our users can learn how to define their particle size in order to have their simulations running as efficiently as possible.

A common scenario in DEM modelling is to have a material with a size distribution where most of the range can be simulated at the real scale, mining rock/minerals like materials, for example. These naturally occuring materials usually have a certain amount of fines (typically small compared to the full size distribution) as a result of having been processed. Translating those fines into a DEM model at real scale can often push the computational requirement at no real benefit. This scenario can be avoided by discarding the lower end of the real size distribution. How much to discard? It would depend on the hardware available and time to be afforded waiting. This is called “Curtailing a size distribution”. Three things must be taken into account when curtailing:

1- Be mindful of gaps in the equipment where particle material can go through

2- Calibrate the material model once the size has been curtailed so that fines are taken into account in the calibrated response via frictions or other parameters

3- Note that analysis on the curtailed sizes will not be available

A good exmaple of a system with a curtailed size distribution can be seen here:

https://community.altair.com/discussion/61379/modelling-brake-dust-in-edem

Following the brake dust example above, If coupling to CFD is required and particle sizes are larger in simulation than in reality, make sure scaling parameters are applied to correctly calculate buoyancy and drag forces as shown in this post:

https://community.altair.com/discussion/36330/edem-api-particle-body-force-field-data-coupling-choice-of-drag-models-with-lift-force

Following on with small particles, it is fair to say that in most cases real sizes cannot be used in DEM modelling when the quantity of material to be simulated is significant. For those cases Meso-scopic modelling is recommended. The objective of this methodology is to capture the complex behaviours of bulk materials made out of micron sized particles (powders and soils) by representing them with compurtationally efficient sizes and advanced physics:

MESO-SCOPIC APPROACH FOR POWDERS AND SOILS:

https://community.altair.com/discussion/40299/webinar-meso-scopic-modelling-of-powders-and-soils-in-edem

The Meso-scopic approach has been widely validated in literature. In the following link you will find an Additive manufacturing case in industry where EDEM was used together with Meso-scopic modelling:

https://community.altair.com/discussion/33682/simulating-a-powder-dosing-system-in-additive-manufacturing-with-edem-material-model-calibration-approach-and-results

In general, if you are working with bulk powders and need to go down the meso-scopic route, EDEM’s unique powders database is the best starting point:

https://community.altair.com/discussion/32951/simple-and-efficient-modelling-of-powders-with-the-edem-powders-database