The Metaseam research project set out to create a process‑specific software demonstrator that can virtually predict laser‑beam welding parameters for a desired weld outcome, and conversely predict weld results from a given set of parameters. The core of the solution is a statistical model built on principal component analysis (PCA). To support this, the team first performed an extensive data acquisition phase, gathering all relevant process, laser, sample preparation, imaging, and weld‑measurement data from previous experiments and new measurements. The collected information was then structured into a hybrid database that integrates diverse data types and sources, enabling the extraction of multivariate relationships between image data of line welds and the corresponding process parameters.

Using the enriched database, the researchers developed a set of PCA‑based models that capture the underlying correlations between laser settings, material properties, and weld characteristics. These models were implemented in a prototype tool that allows users to input either a target weld result or a set of process parameters and receive a predicted counterpart. The prototype also visualises the principal components, giving insight into which variables most strongly influence the outcome. A comprehensive validation campaign was carried out to assess the accuracy and practical applicability of the models. During validation, the models were refined to improve both computational efficiency and predictive precision, ensuring that the final demonstrator meets the needs of industrial users.

The database itself contains a wide range of parameters, including laser power, wavelength, spot diameter, focus position, scan speed, shielding gas flow, and detailed material characteristics such as composition, thickness, and required penetration depth. Imaging parameters such as camera resolution, illumination type, and image‑analysis metrics (e.g., weld width, height, and surface roughness) are also stored, allowing the models to leverage high‑resolution visual data. The continuous adaptation of the database throughout the project added new tables for process data, measurement setups, and image‑analysis results, thereby enhancing the fidelity of the predictive models.

Collaboration among the three research institutions was essential to the project’s success. Fraunhofer Institute for Tool Machining and Forming Technology (IWU) led the data acquisition and database design, ensuring that the experimental setup and measurement protocols were robust and reproducible. The Society for the Promotion of Applied Informatics (GFaI) focused on the statistical modelling and development of the prototype tool, translating the PCA framework into a user‑friendly software demonstrator. Fraunhofer Institute for Materials and Radiation Technology (IWS) contributed expertise in laser‑beam welding physics and material science, providing critical insights into the interpretation of weld‑quality metrics and the selection of relevant process parameters. The project ran from 1 December 2021 to 29 February 2024 under the auspices of the Innovation Fund (project number 22211 BG), which financed the research and facilitated the interdisciplinary exchange of knowledge. Together, the consortium achieved a fully functional demonstrator that lays a solid foundation for future industrial applications of data‑driven welding process optimisation.