A thorough analysis of the requirements for the use cases (UCs) has been completed through collaboration with end users and relevant partners. Currently, data collection is underway to conduct a sustainability analysis of the current processes for all three UCs.
The project achieved successful ultrashort pulse (USP) laser machining of the major geometrical features within the specified tolerances in all three use cases (UCs) using a standard single beam machine which is the first of three stages for the process development. Additionally, a complete laser-based process chain (ablation, cleaning, polishing) for a carbide punch was demonstrated with 87% lower energy consumption than the state-of-the-art process chain.
An adapted training architecture for diffactive neural networks (DNNs) was utilized to calculate the phase masks of two cascaded spatial light modulators (SLMs) with arbitrary amplitude and phase distribution, within the physical limits. In laboratory experiments conducted with a low-power setup, the project successfully demonstrated high-quality beam shaping with both a single SLM and two cascaded SLMs. This achievement was recognized with the Best Student Paper award at the IODC 2023. A prototype is currently being built for a dual SLM module with a high-power ultrashort pulse (USP) source. For this setup, beam steering, manipulated angle of incidence and RMS deviation can be determined in ray optics simulation.
Regarding the development of the quality sensor, a comprehensive analysis of the use cases (UCs) has been conducted, focusing on the perspective of the quality sensor. Specifications for the sensor, including hardware and software requirements, have been defined. Currently, the project is in the process of integrating all the necessary elements for the successful implementation of the quality sensor into the system.
As part of the key objective to develop ML algorithms for real-time laser beam tailoring and process strategies, significant progress has been made in the project. This includes the automation of in-line sensor inputs for ML algorithms and the development of an initial ML model for surface roughness prediction based on sensor data, paving the way for real-time algorithm development.
Progress has been made towards achieving the key objective of "Edge-based in-line feedback process control based on real-time optical sensor data." A platform for real-time acquisition and ML-based analysis has been selected. Furthermore, necessary physical interfaces have been developed and real-time acquisition has been implemented and tested.
The development of the laser machine is on track. The selection of the laser source specifications ensures compatibility with the UC requirements. The determination of the physical size and axes configuration is nearly finalized. The machine will also be capable of accomodate fast rotating rollers for one UC. Additionally, the consortium is actively working on defining interfaces (mechanical, electrical, optical, PC etc.).
