Neural networks for measurement and instrumentation in laser processing

Laser processing is in general a complex process, requiring a lot of knowledge and experience for introducing and maintaining it in industry. This "expert knowledge threshold" obstructs the acceptance of laser technology for new applications. Introduction of process monitoring techniques in combination with sophisticated data analysis tools and artificial intelligence has opened new options to add self-tuning capabilities and closed loop feedback control to laser processing equipment. Some very interesting work has been done in recent years by using soft computing techniques to reach a new level of equipment performance in the field of laser material processing. Advances have been obtained for different types of laser processes, ranging from heavy industry seam welding in shipyard building down to automotive, laser cutting of metal sheets and micro spot welding in the electronics industry. Multi sensor process monitoring systems have been evaluated and their (multi dimensional) outputs related to the process performance through soft computing techniques. Sets of fast sensors are the basic elements to monitor the process from which signal features are extracted and processed by composite traditional/neural-based techniques to perform automatic classification of welded and cut parts. The article presents a comprehensive presentation of the laser processing technology, starting from the basic physics of the process up to a set of industrial applications covering a large range of applications solved by the interaction of traditional processing techniques and neural networks ones.