Current state-of-the-art feedforward controllers of the industrial high-tech machines are limited in performance due to hard-to-model dynamics, such as cable dynamics, frame dynamics, friction, motor driver nonlinearities, time-delays, and couplings between motion axes. The state-of-the art physics-based parametrizations of the feedforward controllers are not sufficient to capture complex dynamics of high-tech machines. This lecture addresses development of a Multiple-input, Multiple-output (MIMO) machine learning (ML)-based feedforward signal generator using Iterative Learning Control (ILC) feedforward data. This ML-based feedforward generator compensates for machine behaviors not captured by the existing feedforward control laws, while being task flexible. The innovative ML-control method is implemented on an ASMPT’s wire bonder and significantly outperforms the existing state-of-the-art physics-based feedforward strategies. Thanks to this method, maximum position errors are lowered by about 63% and the settling times are reduced by over 85%.

Spreker: : Quinten van den Elsen

Company:  Control Engineer, ASMPT

Date: January 12th 2026

Time: 12:02pm          

Location: Teams 

Please send an email to info@dspe.nl if you are interested in following this lecture.

Normally every first monday of the month DSPE organizes for their members a lunch lecture. Each time there is a different topic that a speaker is going to talk about. Due to the holiday period we will organize this lecture in the second week.

Questions can be asked after the 25 minutes presentation. If you want to keep informed about these lunch lectures please send an email to info@dspe.nl