Real-time analysis of inline sensor data during USP-laser machining

Abstract

The challenge of monitoring ultrashort-pulse laser microstructuring lies in the stringent requirements for both spatial and temporal accuracy. Additionally, the monitoring system must not interfere with the processing. This challenge is addressed by employing high-speed off-axis collection of secondary optical emissions. The spatial information is derived from the current processing position, which is recorded synchronously with the emission intensities. A field programmable gate array-based system is used for real-time data collection, synchronization, analysis, and feedback generation. Defects that arise during machining are located as they appear on the workpiece surface, triggering a correction procedure, such as a laser-polishing pass. Furthermore, we compare this method with a data-driven approach using a model that analyzes heatmaps created from photodiode time series and scan positions. A neural network, trained with labels generated by the analytical algorithm and human assistance, detects defects even when the analytical method fails.