Home · DEM Course · Module 2
DEM (Discrete Element Method) Course with BrixDEM

Module 2 · Contact fundamentals

🎬 10 lessons ⏱ 2 h course 📝 5 questions · pass with 4
🔒

This module is still locked

To unlock it, pass the Module 1 quiz with at least 70%.

Go to the quiz · Module 1
Lesson 4 of 10 · 0:39

Friction

Lesson transcript

Now the sideways force. Think of pushing a crate. At first it does not move: friction matches your force. Once you pass the limit, it slides. DEM works the same way, at every contact. The rule has a name: Coulomb friction. The limit is the friction coefficient times the normal force. Below the limit, the contact holds, like a sideways spring. Above it, the contact slides. There is an interesting point here. While the contact lasts, the sideways spring keeps track of how far the particles have shifted. It is the contact's memory. And it is what holds a pile still, without spreading out.

Quiz · Module 2

Check what you learned

5 multiple-choice questions. Get at least 4 right to complete the module. You can try as many times as you like.

✓ You have already passed this module. You may retake the quiz to review.
1.Why does the cost of contact search grow linearly with the number of particles?
2.How does the Hertz model differ from the linear model?
3.In Coulomb friction, when does the contact start to slide?
4.Which rolling resistance model is best suited to heaps and material at rest?
5.What happens when Young's modulus is reduced a hundredfold?
Transcripts of every lesson in this module

1. What happens in a contact

Two particles meet, and everything is settled in less than a thousandth of a second. This module opens that instant in slow motion. First, how the program finds who touches whom. Then, what forces arise at that contact. And finally, why the time step has to be so small. Once you understand this, you know what each parameter does.

2. Who touches whom

The first stage looks easy, but it is not. Comparing each particle with all the others would be far too expensive. With one million particles, that is half a trillion pairs, at every step. The way out is a grid. The domain is split into cells, and each particle only looks at the neighboring cells. Instead of millions of comparisons, there are only a few per particle. That is why the cost grows in a straight line with the number of particles. And there is one more trick. From one step to the next, particles barely move. So the program builds a list of neighbors, with a small margin. And reuses that list for many steps. It is one of the reasons to keep Verlet always on.

3. Normal force

The normal force is the one that pushes one particle away from the other. It comes from the overlap, which you already know. Remember: it is not one sphere really entering the other. It is a way of representing the small deformation at the contact point. BrixDEM has two models. In the linear one, the spring is constant: double the overlap, double the force. It is simple and predictable. In the Hertz model, the spring stiffens as it compresses, just like real elastic spheres. In practice, start with linear, and use Hertz when you need more fidelity. A spring alone would make the particle bounce forever. That is why there is a damper, which removes energy during the contact. You don't set that damper directly. You enter the restitution, and the program computes the rest. At zero point nine, the ball bounces a lot. At zero point two, it almost stops. One more detail. At the end of the contact, the damper could create an attractive force, as if the particles were sticking. That is not physical, and the program sets that force to zero. Sticking is the job of cohesion, which you will see in Module 4.

4. Friction

Now the sideways force. Think of pushing a crate. At first it does not move: friction matches your force. Once you pass the limit, it slides. DEM works the same way, at every contact. The rule has a name: Coulomb friction. The limit is the friction coefficient times the normal force. Below the limit, the contact holds, like a sideways spring. Above it, the contact slides. There is an interesting point here. While the contact lasts, the sideways spring keeps track of how far the particles have shifted. It is the contact's memory. And it is what holds a pile still, without spreading out.

5. Rolling and torsion

Perfect spheres roll with almost no resistance. Real grains have shape, and their contact is a small area that resists turning. That is why rolling resistance exists: a brake you add at each contact. BrixDEM offers three models. CDT applies a constant torque, and it is the default. It is good for flowing material. EPSD works like a spring with a limit, and it is better for piles and material at rest. Viscous only damps the rotation: it does not hold a pile. There is also torsion, when one particle spins like a top on another. The effect is usually smaller, and you switch it on when calibration calls for it.

6. Walls and mesh

How does the program handle a wall? With the same contact physics as between particles. The difference is that the wall has infinite mass: it does not move, even when it takes a hit. A wall is a mesh of triangles. A particle can touch a triangle's face, its edge or its vertex, and the program handles each case. That is why a mesh with distorted triangles or flipped normals can cause trouble.

7. Time step

Now the most important topic of the module. The solver advances in small steps, and a collision has to fit inside several of them. If the step is too large, it skips the start of the contact, the particle sinks deep into the other, and the simulation blows up. The shock wave crosses the particle very fast, and the step has to keep up. The smaller the particle, the smaller the step. The stiffer the material, the smaller the step. And the lighter it is, the smaller as well. That leads to a classic trick. With Young's modulus a hundred times lower, the step gets ten times larger. The price is that particles get softer, and the overlap grows. The rule is to keep it below one percent of the radius. In practice, BrixDEM computes it with the Rayleigh criterion. Start with a factor of zero point two. And if there is cohesion, go down to zero point one.

8. How to know the result is reliable

When someone asks whether the simulation is right, there are really three questions. Verify: does the program solve the equations correctly? Calibrate: do the parameters represent the material? Validate: does it predict something that was not used in the fitting? You can also test this yourself. Simulate the case with one time step. Simulate it again with half of that step. If the result changes a lot, the step was too large.

9. When the case diverges

If the simulation diverges, follow this order. First, cut the time step in half. Then, check whether particles are being born inside the wall. Check the inlet velocity. And look at the mesh.

10. Parameter recipe

To close, a recipe for any new case. Start with the linear model. Use the Rayleigh time step, with a factor of zero point two. Measure restitution, friction and rolling, as in Module 5. And check that the overlap stays below one percent of the radius. With that, you already know what each parameter does.