Advances in Data-driven Optimization of Parametric and Non-parametric Feedforward Control Designs with Industrial Applications

Abstract

The performance of many industrial control systems is determined to a large extent by the quality of both setpoint and disturbance feedforward signals. The quality that is required for a high tracking performance is generally not achieved when the controller parameters are determined on the basis of a detailed model of the plant dynamics or manual tuning. This chapter shows that the optimization of the controller parameters by iterative trials, i.e., data-driven, in both parametric and non-parametric feedforward control structures avoids the need for a detailed model of the plant dynamics, achieves optimal controller parameter values, and allows for the adaptation to possible variations in the plant dynamics. Two industrial applications highlight the large benefits of the data-driven optimization approach. The optimization of the feedforward controller parameters in a wafer scanner application leads to extremely short settling times and higher productivity. The optimization of the current amplifier setpoints in a digital light projection (DLP) application leads to nearly constant color rendering performances of the projection system in spite of large changes in the lamp dynamics over its life span.