HOW TO CHOOSE THE RIGHT PARTICLE PROPERTIES FOR DEM SIMULATION
PARTICLE SHAPE: SPHERES VS. POLYGONS
DEM simulations start with shape. Spheres are the default. They compute fast because collisions only need radius checks. But real particles aren’t spheres—grains, tablets, and pellets have edges. Polygonal models capture those edges, but they cost you. Each collision check now involves face normals, vertex distances, and contact points. If your material is nearly round—glass beads, plastic pellets—stick with spheres. If it’s angular—crushed rock, pharmaceutical tablets—polygons are non-negotiable.
Particle shape directly impacts flow behavior. Spheres roll too easily; polygons interlock. A 10% shape deviation can flip your angle of repose from 25° to 40°. Measure your real particles with a 3D scanner. If the sphericity (ratio of surface area to volume) is above 0.9, spheres will save you weeks of runtime. Below 0.8, polygons are the only way to trust your results.
PARTICLE SIZE DISTRIBUTION: MONO VS. POLYDISPERSE
Single-size particles simplify your model. Every collision uses the same mass, same stiffness. But real processes rarely have uniform particles. A crusher feed might span 10 mm to 100 mm. Ignoring that range means your simulation will under-predict segregation and over-predict flowability.
Polydisperse models add complexity. You need a representative sample—at least 500 particles per size bin. Use sieve data or laser diffraction to build your distribution. Then decide: discrete bins or continuous curve. Discrete bins are easier to code; continuous curves match reality better. If your process depends on fines (below 1 mm), continuous is mandatory. Fines clog, fluidize, and dominate surface area. A 5% fines fraction can drop your bulk density by 20%.
PARTICLE STIFFNESS: REAL VALUES OR CONVERGENCE HACKS
Stiffness controls overlap during collisions. Too soft, particles interpenetrate like ghosts. Too stiff, your time step shrinks to nanoseconds. Real materials have Young’s moduli from 1 MPa (soft plastics) to 200 GPa (steel). Plugging those values into DEM gives you time steps around 1e-8 seconds. A 1-second simulation now needs 100 million steps.
Most users cheat. They drop stiffness to 1e6 Pa and scale damping to keep restitution coefficients realistic. This works for flow-dominated systems—hoppers, conveyors. But if you’re simulating compaction or milling, stiffness matters. A 10x stiffness increase can double your contact forces. Measure your material’s Young’s modulus with a nanoindenter or use literature values. If you must scale, keep the ratio of normal to shear stiffness constant. Never scale one without the other.
COEFFICIENT OF RESTITUTION: ENERGY LOSS OR ENERGY CHAOS
Restitution tells you how bouncy your particles are. 1.0 means perfect bounce; 0.0 means dead stop. Real values sit between 0.2 (sand) and 0.9 (steel balls). High restitution keeps energy in the system, leading to fluid-like behavior. Low restitution kills motion fast, mimicking damp materials.
Measure restitution with a drop test. Film a particle falling onto a flat plate of the same material. Track its rebound height. Do this 10 times; average the results. If your particles are irregular, use a rotating drum test instead. Restitution isn’t constant—it drops with impact velocity. Most DEM codes ignore this. If your process has high-speed impacts (mills, chutes), use a velocity-dependent model. Otherwise, a single value is fine.
FRICTION: STATIC VS. ROLLING RESISTANCE
Friction stops particles from sliding. Static friction (μs) sets the threshold; rolling friction (μr) resists rotation. Both are critical. A μs of 0.3 gives you a 17° angle of repose. Bump it to 0.6, and your pile stands at 30°. Rolling friction is sneakier. It mimics shape effects without polygons. A μr of 0.1 can turn spheres into cubes in terms of flow behavior.
Measure static friction with a tilt table. Place two particles on a plate, tilt until one slides. Rolling friction is harder—use a drum test or a shear cell. If you’re using spheres, rolling friction is your only way to approximate real shape. Without it, your particles will flow like marbles. For polygons, rolling friction is less critical but still needed to prevent unrealistic rotation.
PARTICLE DENSITY: BULK VS. TRUE
Density seems simple. Look it up, plug it in. But bulk density and true density are different beasts. True density is mass per volume of solid DEM Simulation . Bulk density includes voids. A simulation with true density will over-predict mass flow rates by 30-50%. Always use bulk density for your particle properties.
Measure bulk density with a graduated cylinder. Fill, tap, weigh. Do this three times; average. If your particles are porous (catalysts, pharmaceutical granules), true density won’t match. Use helium pycnometry for true density, then adjust your particle mass to hit the measured bulk density. Never assume literature values. A 5% density error can shift your discharge rate by 10%.
CONTACT MODELS: LINEAR SPRING-DASHPOT VS. HERTZ-MINDLIN
The contact model defines how particles interact. Linear spring-dashpot is the simplest. A spring for stiffness, a dashpot for damping. It’s fast, stable, and wrong for most materials. Hertz-Mindlin is nonlinear. Stiffness scales with overlap; damping depends on relative velocity. It’s slower but matches real physics.
Use linear only for quick feasibility studies. Hertz-Mindlin is mandatory for quantitative results. If your particles are soft (polymers, food), add a plastic deformation model. If they’re brittle (glass, ceramics), include a breakage model. Never mix models in the same simulation. A single linear contact in a Hertz-Mindlin system will act like a black hole, sucking energy out of your simulation.
TIME STEP: STABILITY VS. SPEED
Time step ties everything together. Too large, particles tunnel through each other. Too small, your simulation takes weeks. The rule of thumb: time step = 20% of the Rayleigh time. For a 1 mm particle with 1e6 Pa stiffness, that’s 1e





