Development of robot deflection compensation for improved machining accuracy

The project described in this document was undertaken to demonstrate the feasibility of compensating for robot deflection during machining operations. A Cincinnati Milacron T3-776 robot has been used to develop and demonstrate the modeling and control technology required to compensate (or reject) the robot deflection that would otherwise occur during a machining operation. This technology has been demonstrated while cutting curved aluminum parts with a high-speed milling cutter.; Several subtasks have been completed in preparation for the demonstrations. The original control computer supplied with the robot was replaced with a modular controller developed for this purpose. In addition, models of the robot's kinematic structure and compliance properties were developed for use by the deflection compensating control algorithm.; This work focuses on the additional system modeling and control development required to demonstrate deflection compensation and to initiate continued growth of the technology. High-fidelity models of individual system components (individual axes, power amplification, and the cutting process) were developed and verified for use in control synthesis and numerical simulation. Nonlinear friction has been shown to have significant impact on robot response and is included in the high-fidelity models of robot axes.; Two separate control efforts are presented. The first makes use of a linearized model of the robot system which includes a position dependent compliance matrix to model robot deflection. The poles of the linearized and discretized system are placed inside the unit circle of the Z domain with linear state feedback. The second control approach was motivated by the inability to directly apply the first approach without additional compensation for nonlinear effects such as the sticking friction. The second approach is an ad hoc approach similar to force control algorithms found in the literature with the addition of an explicit model of robot compliance. The same position dependent compliance matrix mentioned above is used with this algorithm. This algorithm has been implemented and demonstrated to be effective in compensating for bandwidth limited end-effector loading.