The FA4.0 project, funded by the German Federal Ministry of Education and Research (BMBF) under the grant number 16ME0112 and coordinated by VDI/VDE Innovation + Technik GmbH, ran from 2020 to 2023 with adjustments for the COVID‑19 pandemic. During this period the consortium held several workshops, including the CAM‑Workshop in Halle in 2022 and 2023 and the ESREF conferences in Berlin and Toulouse from 2020 to 2023, to exchange progress and refine the research agenda. The project was organised into five work packages covering standardised interfaces, data management, artificial‑intelligence algorithms, advanced instrumentation, and integrated workflows. Key partners were Bosch, Infineon, ST, Matworks, KERN, TESCAN, Cyber Technologies, Excillum, and the University of Stuttgart, each contributing specialised expertise and equipment.
In Work Package 2 the consortium delivered a second‑generation sample holder that successfully enabled a seamless workflow from light microscopy to scanning electron microscopy, demonstrated at the Grenoble meeting in March 2023. A common data header was finalised jointly with Infineon and ST, and a menu‑driven tool called Headermaker was released to create and edit this header. The header, once implemented by machine and tool manufacturers, allows full traceability of the analysis process and supports the creation of digital twins for sample and equipment parameters, leveraging ontology concepts developed at the University of Stuttgart.
Work Package 3 focused on machine‑learning‑based image and signal analysis. The AI‑driven image evaluation program for IMC‑coverage was expanded with training data for copper and silver, packaged as an executable for all analysts. The tool was rolled out across Bosch’s automotive analysis department, delivering a minimum ten‑fold reduction in image‑processing time. Parallel efforts advanced SAM, infrared, and cratering test image analysis. In signal analysis, the evaluation of 5,000 sensor elements using z‑scans was completed, with physical insights fed back into the FA2IR project.
Work Package 4 introduced several advanced instruments. An acoustic microscope incorporating SAFT and AI analysis was optimised for internal focus, enabling simulations with PVA software and improving reconstruction accuracy. A local milling tool for decapsulation was developed through collaboration with KERN, converting in‑house equipment into a 3D‑capable system. A FIB/REM system was equipped with the new data header, enhancing decision‑making in equipment procurement. Cyber Technologies supplied measurement systems for thermally induced bending, offering high automation and task‑specific adaptability. Excillum’s ultra‑high‑resolution X‑ray scanners were shown to provide superior detail and operate at roughly twice the speed of Bosch’s scanners, a performance advantage that will inform future device evaluations.
Finally, Work Package 5 integrated the developed tools into end‑to‑end workflows. The acoustic microscope, FIB/REM system, and X‑ray scanner were combined with digital‑twins‑based data analysis and defect catalogues to create comprehensive defect‑localisation procedures across mechanical, laser, and FIB cross‑sections. The project’s outcomes, including the standardized data header, AI‑enhanced image and signal processing, and upgraded instrumentation, are expected to set new benchmarks for defect analysis in the semiconductor and automotive industries.