Simultaneously to the rapidly growing number of industrial applications in the field of laser machining in many sectors, an increasing demand for automated process quality monitoring systems can be found. Industrially-used process monitoring systems are usually based on optical sensors evaluating the radiation that is emitted from the process. Such mechanisms are highly reliable for detecting process irregularities, but there are still disadvantages that currently cannot be overcome. Two fundamental problems can be criticized: firstly, it is almost impossible to classify faults. Secondly, current systems need inflexible reference signals which are very time-consuming to create. The second problem is particularly impractical since references become invalid as soon as parameters, such as the process duration, are changed.
To overcome such restrictions it is necessary to get more infonnation from the process. Therefore, an approach with additional sensors and advanced evaluation methods was investigated. The goal was to create flexible algorithms that can be used for the observation of laser welding as well as laser cutting. Experimental working heads for both processes were designed, based on industrially-used working heads. To achieve a reference-free signal evaluation that provides more information about the process, sophisticated algorithms were developed. Incoming signals are processed using statistical methods. Characteristic faults can be determined by combining results from the single methods. Fuzzy membership fimctions are used to fmally generate a standardized overall quality rating.