The use of lasers as a photonic tool is industrially established and has led to a transformation in which conventional manufacturing methods are being replaced by laser-based approaches. However, the wide range of process conditions that are possible require a considerable amount of effort to be spent on the iterative identification of process-stable and speed-optimized parameters. In particular, the generation of surface topographies with features in the nano- and micrometer range is time-consuming, especially when using disruptive laser interference-based technologies, such as direct laser interference patterning (DLIP). In this area, it has already been successfully demonstrated that AI-based approaches can significantly accelerate process development and the technological exploitation of functionalized surfaces (e.g. self-cleaning surfaces modelled on the lotus effect) [1]. However, this can only be expected in a comprehensive solution with process-integrated data collection, in which the self-learning production control system can automatically adapt to the changing context based on previously generated real-time process data in order to maintain optimal quality.
This challenge is now being addressed by a self-developed state of the art laser processing system (AI-Testbench), which integrates micro material processing with artificial intelligence and machine learning. The AI-Testbench has one measurement and two machining platforms, connected by a precision axis-system. The processing stages include a polygon scanner, a flexible beam former, a 3D scan system and a leading-edge DLIP module for laser interference texturing. Beam sources include a high-power ultra-short pulsed laser (P = 300W, tp  = 1.6 ps, Ep = 3 mJ) and a short-pulsed fiber laser system (P = 200W, variable pulse duration and shape). During material processing all process data is automatically captured, such as laser power, intensity distribution, visual and acoustic emissions from the process, reflectivity, emission of secondary radiation etc. Fabricated structures can be evaluated for a wide range of surface and functional properties using multiple measurement systems, like gloss of the surface, ablation depth, resulting roughness and contour accuracy. In addition to a white light interferometer, a hyperspectral camera is installed at the measurement platform. For real-time process monitoring, an FPGA-supported audio and video evaluation is included, which can feed back the smallest changes that occur during machining into the process. For data acquisition and monitoring, the system is fully digitized via an integration platform to easily connect the different interfaces.

[1]Tobias Steege et. al., „Predicting laser-induced functional surface textures: A comparison of random forest and neural networks“, presented at the EPoSS & EPSI Annual Forum 2020, 2020.