Propulsion comes from acceleration of a reaction mass.
In this case the grinding wheel serves two purposes:
- separate small bits of the workpiece from bulk at a slow and roughly uniform rate
- accelerate those bits by mechanical friction, in a similar way that a tennis ball launcher uses a matched pair of counter-propagating spinning wheels to shoot a box full of tennis balls one at at time in a controlled direction and speed.
As @Muze points out using a matched pair of counter-propagating spinning wheels would be important in spaceflight as well.
Step 1: requires a large amount of work and there's no reason to do that in space. You can produce the particle on the ground, so your propellant "tank" would be a feed system dispensing pellets or powder. They could be suspended in a fluid for easier feed and to avoid electrostatic clumping.
If you have to produce it in space, for example if you are reusing your lower stage as a reaction mass Horace style (Monty Python reference), then you can grind or otherwise form at a lower speed first with a separate wheel. Particles could be remelted in order to make them spherical, and then size-sorted for the following step.
Step 2: would conceptually be accomplished with a mechanism similar to that of a tennis ball launcher. Two counter-propagating wheels with the particle feed introduced in the small gap between the two wheels' surfaces. The particles would have to be monodipserse, meaning all of a fairly uniform size, and slightly smaller than the gap for good friction. Either the particles or the wheels would have to be elastically compressible enough so that there's a good grip for acceleration, and yet the surfaces should not be readily damaged during the process.
You could also very slightly incline the wheels such that if you had a sorted range of particle sizes they can all be introduced at the appropriate gap width.
However, as @Greg points out and then @ Thorondor demonstrates quantitatively, getting your wheels to spin at mach 10 or faster (for a (mass-)specific impulse or Isp of say 300 or so) is a real materials problem.
One possibility for propellant would be a maximal concentration (essentially HCP) liquid suspension of latex or polymer spheres which can be obtained highly monodisperse with sufficient money. If you don't have a lot of room for propellant, then perhaps metallic or metal oxide or nitride nanospheres can be produced by pyrolysis.
GIF from Tennis ball machine DIY - part 1 a better view of a similar mechanism can be seen in Tennis Tutor Ball Machine mechanics in operation.