Inverse dynamics toolpath compensation for CNC machines based on model predictive control

Abstract

The use of model predictive control (MPC) as a form of inverse dynamics compensation for multi–axis CNC machines, to subdue the inaccuracies incurred by axis inertia and damping, is investigated by both simulation studies and experimental performance analysis using a 3–axis milling machine governed by an open–architecture software controller. The results indicate that MPC is a viable tool for inverse dynamics compensation with a controller sampling frequency f = 1024 Hz running on a 500-MHz processor, with only modest prediction horizons offering excellent performance in terms of feedrate accuracy and contour error suppression. Unlike inverse dynamics schemes based upon linear time–invariant dynamic models, the MPC scheme provides the flexibility to compensate for nonlinear physical effects such as backlash in the machine axes and hard constraints on axis accelerations imposed by motor torque constraint.