Mesh convergence - Determining sphere radius

Hello Justin,

> I am moving a tool down into a box of balls and then rotating the tool. I am having the Torquerecorder output the torque. I have a few questions. I have a background in FEA and still learning DEM.
> Is DEM even capable of analyzing my tool this way?

for better answer, please describe your problem more, e.g. also with some illustrative images. But as I understand it, yes, DEM is capable for analysis of such problem.

> Mesh convergence - In FEA you can do a mesh convergence study to determine a small enough mesh. Is something like this done with DEM? I have ran 8 different simulations with radius' ranging from 1 to 4.5 (in intervals of .5). I have received completely different results, by factors of almost 100.

There are two approaches, what particles and their size could be.
Either it represents physical grains (e.g. sand or gravel). In that case, you simply simulate physically different problem and different results are expected. Just imagine a real tool mixing 1 mm grains and 5 mm grains. Also, some convergence analysis does not make sense, since the problem is different for each size.
Or the particles represent an artificial discretization of continuous material, and in that case some convergence study makes sense.

> What is the best method of determining a good radius for the spheres?

Depend on what type the spheres are.
If they represent physical grains, their radius is determined "by default".
If they are artificial discretization, the radius can be determined according to desired resolution, computation costs etc. In general, it is very individual :-)

> Is there a good rule of thumb when trying to calibrate your material?

To answer, please provide more info what you want to calibrate.

> Also, I read if you increase mass of the sphere you can decrease the time it takes to complete. Is there any rule of thumb on this? If I increase the mass, do I decrease something else to still receive accurate results?

DEM uses explicit time integration with critical time step of mass-spring system proportional to sqrt(mass/stiffness). If you increase mass, you increase the critical time step, so you decrease time of simulation. On the other hand, you change the physics of your problem, so if you "still receive accurate results" depends very much on the problem, but in general not.
This is similar to explicit FEM dynamics. There you can also increase material density to increase time step, but changes the problem physics. In some cases it would be ok, while in other cases it would just be stupid..

cheers
Jan