A methodological approach to multisensor classification for innovative laser material processing units

Online quality detection and online laser beam control are important research topics to improve the overall quality of presentday laser beam material processing units. In both cases innovative units are being studied where the state is monitored by a set of heterogeneous in-process sensors conveying a large amount of information. However, low experimental reproducibility, lack of dominion knowledge and high costs greatly limit our ability to find an optimal solution. In this paper we propose a methodology to guide the engineer's design choices towards an optimal implementation of the inductive classifier.