Data-driven Parameters Tuning for Predictive Performance Improvement of Wire Bonder Multi-body Model

This report describes work performed during SWI 2023 at the Univer-sity of Groningen in relation with Problem 1 posed by the companyASMPT.ASMPT makes a very large number of different machines for manufac-turing of electronic devices. They have detailed simulation software ofone of these machines and they compare the results of this with phys-ical experimental results. There is a significant difference between thesimulated and measured data, and it is the goal of this work to studyhow to estimate the parameters in the simulation model using the ex-perimentally measured frequency response.First, two toy models are studied to understand the challenges of pa-rameter estimation in the frequency domain. Later, optimization meth-ods are applied. Several different approaches of reducing the dimen-sionality of the parameter space are explored, including determiningthe parameter sensitivity. A suggestion for increasing the detail of themodel, specifically related to the machine base, is also outlined.In the summary, we supply a discussion of the key insights we gainedduring the week.