TY - CHAP

T1 - Motion and Vibration Control

T2 - Applying Iterative Learning Control for Accuracy Improvement of an Electromagnetically Actuated Punch

AU - Dagen, M.

AU - Abdellatif, H.

AU - Heimann, Bodo

PY - 2008

Y1 - 2008

N2 - Many industrial processes are characterized by a cyclic mode of operation. Thus, based on an identical initial condition the process performs the same task in a finite time span over lots of repetitions. The challenge is to follow a desired trajectory as good as possible. Conventional non-learning controls use the error in the time domain only and therefore cannot compensate tracking errors excited by deterministic disturbances or unconsidered dynamics of the system. Hence, a non-learning controller yields the same tracking error at each iteration. This paper presents an application of Iterative Learning Control (ILC) for optimizing the cutting process of an electromagnetically actuated punch (EAP). In contrast to mechanical presses, with the EAP it is possible to change the ram's kinematics freely and to optimize it online. During the contact of the ram with the work piece, high transient forces are excited and deteriorate the positioning accuracy of the ram. By using a Sliding-Mode-Control it is not possible to compensate this. Thanks to the cyclic nature of the cutting process, we apply ILC in order to increase the accuracy of the ram. In this work we present a comparison study of two linear approaches. The first one consists in a filtered and phase lead compensated integral learning. In contrast, the second approach exploits explicit knowledge of the system's experimentally identified transfer function and performs a contraction mapping during the learning process. The experimental results show that both algorithms are capable to reduce the positioning error and to increase the accuracy of the system, even at high dynamics.

AB - Many industrial processes are characterized by a cyclic mode of operation. Thus, based on an identical initial condition the process performs the same task in a finite time span over lots of repetitions. The challenge is to follow a desired trajectory as good as possible. Conventional non-learning controls use the error in the time domain only and therefore cannot compensate tracking errors excited by deterministic disturbances or unconsidered dynamics of the system. Hence, a non-learning controller yields the same tracking error at each iteration. This paper presents an application of Iterative Learning Control (ILC) for optimizing the cutting process of an electromagnetically actuated punch (EAP). In contrast to mechanical presses, with the EAP it is possible to change the ram's kinematics freely and to optimize it online. During the contact of the ram with the work piece, high transient forces are excited and deteriorate the positioning accuracy of the ram. By using a Sliding-Mode-Control it is not possible to compensate this. Thanks to the cyclic nature of the cutting process, we apply ILC in order to increase the accuracy of the ram. In this work we present a comparison study of two linear approaches. The first one consists in a filtered and phase lead compensated integral learning. In contrast, the second approach exploits explicit knowledge of the system's experimentally identified transfer function and performs a contraction mapping during the learning process. The experimental results show that both algorithms are capable to reduce the positioning error and to increase the accuracy of the system, even at high dynamics.

M3 - Contribution to book/anthology

VL - 9

SP - 41

EP - 51

BT - Proc. of the 9th International Conference on Motion and Vibration Control

A2 - Ulbrich, H.

A2 - Ginzinger, L.

PB - Springer Netherlands

ER -