Use spindle torque to control feedrate?

[Jonathan]

The usual way to calculate a feedrate for the tool and machine you're using is to consider the properties of the material and tool you're using and set the feedrate to a value which you're confident wont put enough force on the tool to break it. This is a somewhat conservative method, as you have to reduce the feedrate such that the tool survives the 'bad bits' of the toolpath. So why not measure the force on the tool as it is cutting and adjust the feedrate to maintain the force at a value just below the breaking force of the tool? Measuring the force is difficult, so lets do the next best thing and control the torque at a spindle (easy with vector control).

Now we just need to derive a suitable control system - the input is a torque reference (just under breaking torque of tool), feeback is current spindle torque output and the output is the feedrate. The controller will continuously adjust the feedrate to always run the tool at the highest torque (force?) it can tolerate.

Some obvious obstacles with this are the bandwidth of the system (i.e. can we change the feedrate fast enough), the linearity of the torque vs feedrate characteristic of the tool, and the software side of it. I don't think the software will be a problem however, since LinuxCNC has been used to adjust feedrates to get constant current in wire EDM machines and this isn't much different from a control point of view. Some simulation would help here, but speculatively a PID controller might be sufficient. Could potentially add feedforward term based on calculating tool engagement and predicting torque from that, but that's a bit tricky. Although the response time of the mechanical system is slow, the torque response time of servo spindle motors can be very fast (few ms) so our feedback signal is good...Could perhaps add additional control loop to adjust the spindle speed if the torque limit is about to be exceeded - e.g. if accelerating the spindle up reduces the force on the tool, then it's OK to do that briefly so long as it's not for long enough for too much heat to build up.

There are of course expensive CAM programs that change the feedrate based on tool engagement, which is particularly effective when combined with trochoidal milling. However that's still based on estimating what forces the tool will tolerate. With my proposal we're measuring it, so should be able to run closer to the limit and of course use cheap CAM programs.

Any ideas? When I've got more time I'll have a go at simulating with simulink, or just try it...whichever comes first!