Comparison of two iterative learning concepts applied on a manipulator with cogging force disturbances

In laser welding applications, either the laser beam or the workpiece has to be manipulated such that the seam to weld can be tracked. Gantry type of robots driven by linear direct drive synchronous motors can be applied as manipulator(s). In direct drive permanent magnet motor systems, disturbances due to cogging forces are directly noticeable at the end-effector and influence the performance. In the usual P(I)D controlled set-up of these systems, cogging forces induce disturbances which cause errors in seam-tracking accuracy in such a way that laser welding at high speeds (up to 0.5 m/sec.) do not result in qualitatively good weldings. In here it is shown how Iterative Learning feedforward Controller techniques can be designed and applied to achieve a high precision tracking accuracy. Two different ILC techniques, one using a zero phase error approach and one using H∞ synthesis, are compared. The conclusion is that the ILC using the a-causal zero-phase error approach gives the best results and the ILC using H∞ synthesis suffers from phase errors.