The discrete element method (DEM) is a way of simulating bulk material on a computer by treating it as a very large number of individual particles. Instead of treating coal, ore or grain as a smooth fluid, DEM follows each particle separately. In tiny steps of time, it works out which particles are touching each other or a wall, what forces those contacts create, and how each particle moves as a result. Repeat that millions of times and you see how the whole stream behaves: where it flows freely, where it slows, where it piles up and where it hits hardest.
That makes DEM simulation a practical tool for anyone who designs or runs transfer chutes, conveyors, hoppers, bins and shiploaders. This guide explains how it works, what it is used for, how it differs from discrete event simulation (DES) and CFD, the main DEM software, and what makes results worth trusting.
How DEM simulation works
A DEM model has three parts: the particles, the geometry they move through, and the rules for what happens when things touch.
Particles
Each particle has a size, shape and density. A model usually uses a size distribution from a sieve test. Shapes range from spheres (fast) to clumps of spheres or polyhedra (slower, but closer to angular coal or rock).
Contact models
The contact model turns "these two particles overlap slightly" into a force. The most widely used is Hertz-Mindlin, which treats each contact like a spring and damper: a push-apart force from the overlap, a sideways force from sliding friction, and energy lost in the bounce. Most models add:
- Rolling friction, so round model particles don't roll away forever; it stands in for real, irregular shapes.
- Cohesion, for wet or sticky material. Damp coal, clay-rich ore and fines stick to each other and to steel. Cohesion models (such as JKR) add an attractive force so simulated material can build up on walls and form arches, as the real material does.
- Wall properties: friction and bounce against steel, rubber or ceramic liners.
The time step
DEM moves forward in very small time steps, often millionths of a second. At each step the software finds every contact, calculates the forces and moves every particle. Stiffer particles need shorter steps, so longer run times.
Calibration
Calibration makes the simulated material behave like the real one. Samples are tested in the laboratory, and the model's settings are adjusted until simulated versions of the same tests match. Common tests:
- Angle of repose: the slope a heap of material settles at.
- Shear tests: how much force it takes to make a layer of material slide over itself, which shows internal friction and cohesion.
- Wall friction tests: how material slides on a sample of liner or steel.
- Drawdown or flow tests: how material empties from a small box or funnel.
An uncalibrated model can produce attractive pictures that don't match what happens on site.
What DEM is used for in bulk handling
DEM is used wherever the question is "how will this material move through this equipment?" Common uses:
- Transfer chute design. Testing a new chute, or a change to an existing one, before steel is cut. See our guide to transfer chute flow analysis for the detail.
- Chute blockage. Showing where wet or sticky material slows, builds up and chokes a chute, and testing changes to wall angles, liners or flow aids.
- Liner wear. Mapping where material impacts hardest, so wear-resistant liners go where needed.
- Belt loading and centring. Checking that material lands centrally on the receiving belt, moving in the belt's direction near belt speed.
- Spillage and dust. Showing where the stream breaks up and where fines are likely to escape.
- Hoppers, bins and silos. Checking whether material flows evenly or forms a stagnant zone or a rathole, and whether it arches over the outlet.
- Coal handling plants. Many transfer points, crushers and bunkers, with coal that changes by source and season. Each problem transfer point can be studied with the coals that actually pass through it.
- Shiploader spouts. How material flows at different boom positions and loading rates.
- Conveyor chute and belt conveyor design generally: feed boxes, skirts, impact zones and discharge trajectories.
DEM shows behaviour inside the equipment that is very hard to measure on a running plant.
DEM vs DES vs CFD vs rules of thumb
People often mix up DEM and DES, partly because the names are similar. They answer different questions.
Discrete element modelling vs DES
Discrete event simulation (DES) models a system as a series of events over time: a train is unloaded, a ship berths, a shiploader moves between hatches, a conveyor stops for maintenance. It doesn't model particles. It answers "how many tonnes per day can this terminal move?" or "what happens to ship loading time if one conveyor is down for a shift?"
DEM models what happens inside one piece of equipment over seconds or minutes. It answers "will this chute block with wet coal at this rate?"
They work well together: DES gives the duty a transfer point must handle; DEM shows whether the equipment can.
DEM vs CFD
Computational fluid dynamics (CFD) models air, water and slurry as a continuous flow, which suits dust extraction and pumped slurry. Bulk material in a chute is driven by particle contacts, friction and cohesion, which is what DEM is built for. Some studies couple the two, for example to model dust around a transfer point.
Spreadsheets and rules of thumb
Hand calculations, trajectory methods and rules of thumb are often enough for a familiar material at a familiar rate. DEM earns its place when the material changes (several coals, wet seasons, more fines), throughput rises above design, the layout is new, or a chute keeps blocking and nobody can see why.
DEM software options
This is a vendor-neutral overview, not a recommendation.
Commercial DEM software
- Altair EDEM. One of the most widely known DEM products, which is why people search for "EDEM" by name. Altair describes its GPU and multi-GPU solver engine as based on NVIDIA CUDA, alongside its CPU solver.
- Ansys Rocky. Another leading DEM product. Ansys describes a native multi-GPU solver that spreads a simulation across two or more GPUs in one machine.
Both include calibration tools, non-round particle shapes, wear prediction and CFD coupling.
Open-source DEM software
- Kratos Multiphysics (DEM Application). A C++ framework with a Python interface. Kratos describes itself as OpenMP and MPI parallel, so it runs on CPU cores, from a workstation up to a cluster.
- LIGGGHTS and Yade. Long-standing open-source DEM codes that run on CPUs and are widely used in research.
- DEM-Engine (DEME), from Project Chrono. A GPU DEM solver released under the BSD-3-Clause licence. Its developers describe it as simulating granular material using one or two NVIDIA GPUs, with support for complex particle shapes.
- PhasicFlow. An open-source DEM code under the GPL-3.0 licence, built on the Kokkos library so it can run on multi-core CPUs and CUDA-enabled NVIDIA GPUs.
Open-source tools avoid licence fees but need more scripting, set-up and checking.
CPU vs GPU for DEM
DEM is heavy computing. It's natural to assume more CPU cores will make it faster, but past a point that often isn't true:
- Synchronisation every time step. Every core must finish its share before any particle moves on, so the job waits for the slowest core, thousands of times a second.
- Uneven particle density. Most particles are packed in the stream, and much of the chute is empty. Split by region, some cores get crowded areas and others almost nothing.
- Memory bandwidth. DEM constantly reads and writes positions and forces. Once memory can't feed data faster, extra cores just wait.
- Two-socket penalty. On two-CPU servers, reaching memory attached to the other CPU is slower, so jobs across both sockets can scale worse than expected.
GPUs suit DEM: thousands of simple cores, very fast memory, and contact calculations that are small and alike. Many GPU DEM codes use single precision, which consumer and workstation GPUs handle much faster than double precision; for most bulk handling problems that is accurate enough, but it is worth knowing.
Limits of DEM and what makes results trustworthy
DEM is powerful, but it relies on simplifications. A trustworthy study is open about them.
Particle scaling and coarse-graining
A real chute carries far more particles than any computer can follow, especially fines. So models use larger particles than the real material (particle scaling), or let one model particle stand in for many (coarse-graining). Calibrated at the scale used, this gives useful results. Done carelessly, it can hide the fines-driven sticking behind many blockages.
Reduced stiffness
Real particles are very stiff, which would force extremely short time steps. Most studies lower the stiffness so the model runs in reasonable time. That is standard and generally fine for flow patterns, but it can affect impact forces, so it should be checked against the question asked.
Calibration and validation
The two things that make results believable are:
- Calibration against laboratory tests on the actual material, ideally at more than one moisture level if the site has wet seasons.
- Validation against what the site already knows: where the existing chute builds up, where liners wear through, where spillage happens, and how it behaves at different throughputs.
Throughput records, stoppage logs, liner replacement history and moisture readings let the model of the current chute be checked against reality before it judges a new design. At many sites that information sits in separate systems and spreadsheets, so bringing it together is part of good DEM work.
DEM gives a strong indication of behaviour and is excellent for comparing options. It is not a precise prediction of every particle.
Before and after: comparing the current chute with a modified design
The most useful DEM study is usually a comparison. First the current chute is modelled with calibrated material at real throughputs. If it shows build-up and wear where the site sees them, there is good reason to trust it. Then the modified design is run under the same conditions: same material, same rates, same wet and dry cases.
Side by side, the two show what changed: whether build-up has gone, whether impact has moved off the liners that keep failing, and whether material now lands centrally on the belt. Because both runs share the same assumptions, the comparison is more reliable than either result alone.
Kenomont is publishing before-and-after simulation videos showing this kind of comparison. The first, a transfer chute shown as the current and a proposed design, is on our bulk handling simulation page.
Where Kenomont fits
We have built transfer chute flow simulations ourselves: modelling how material moves through a chute, showing where it builds up or wears, and testing changes in the model before anything is fabricated. The final chute design, fabrication and sign-off stay with the site's engineers.
We work remotely with companies in Australia and across Asia, using drawings, conveyor data and material information supplied by the site, and we can help bring together the throughput, stoppage and maintenance data that makes a simulation worth trusting.
Frequently asked questions
What is the discrete element method?
The discrete element method (DEM) is a computer simulation method that models bulk material as many individual particles. It calculates the contacts between particles and walls, and the motion of every particle, step by step through time.
What is DEM simulation used for?
In bulk handling: transfer chute design, chute blockage, liner wear, belt loading, spillage and dust, hoppers and bins, coal handling plants and shiploader spouts.
What is the difference between DEM and DES?
DEM (discrete element method) models particles moving inside equipment over seconds or minutes. DES (discrete event simulation) models a whole operation as a series of events, such as trucks arriving, ships loading and conveyors stopping, to answer throughput and capacity questions. They are often used together.
Is EDEM the same as DEM?
No. DEM is the method. Altair EDEM is a software product that uses the method, in the same way that Ansys Rocky, Kratos, LIGGGHTS, Yade, DEM-Engine and PhasicFlow do.
Can DEM run on a GPU?
Yes. EDEM and Rocky both have GPU and multi-GPU solvers, and open-source DEM-Engine and PhasicFlow run on NVIDIA GPUs. GPUs suit DEM because the work is many small, similar calculations that need fast memory.
How accurate is DEM?
It depends on calibration and validation. A model calibrated with laboratory tests on the real material, and checked against how the existing equipment behaves on site, gives a reliable picture of flow patterns and is excellent for comparing designs. Simplifications such as particle scaling and reduced stiffness mean it should be read as a strong indication, not an exact prediction.
Can you run a DEM study for a coal handling plant outside Australia?
Yes. We work remotely with companies in Australia and across Asia, using the site's drawings, conveyor data and material information. We run the simulation and report the results; final design and sign-off stay with the site's engineers.
Talk to us about a chute or conveyor problem
If you have a transfer chute that blocks, wears or spills, or you want to test a design change before fabrication, we can help with a DEM study and with the operating data behind it. Email [email protected] or use the form to book a free initial chat, wherever your plant is. Callers in Australia can also ring 1300 470 091.