Settling Time Optimization in Wire Bonder Systems via Extremum-Seeking Control1

Abstract

Adequate tuning of control laws is essential for high positioning accuracy, large system throughput, and reliability in high-end mechatronic and robotic systems. However, a population of such systems generally shows slight variations in dynamic responses due to, e.g., manufacturing tolerances, different disturbance situations, or position-dependent dynamics. Given the time-consuming nature of controller design, even by experienced control engineers, typically just one control law is designed for the whole system population based on worst-case bounds on variations in dynamic responses, resulting in a loss of individual system performance. The main contribution of this paper is the development of an automated controller tuning approach, based on extremum-seeking control, for settling time optimization via individual controller tuning. While other automated controller tuning methods exist, the developed approach allows inclusion of closed-loop stability and robustness constraints based solely on non-parametric frequency-response measurements of open-loop plant dynamics, and therewith directly optimizes transient system performance in a purely data-based manner. The proposed approach has been applied in simulation in an industrial case study for settling time optimization in point-to-point motions of a wire bonder system. In this case study, the effectiveness of the approach has been shown by achieving significant performance increases of 39.4% and 40.6% compared to controllers designed by experienced control engineers using manual loop-shaping techniques and a frequency-based auto-tuner, respectively, without needing manual tuning effort.