Xecs FA2IR

FA2IR builds on the Penta/Euripides project FA4.0, which has demonstrated the use of Artificial Intelligence (AI) algorithms in Failure Analysis (FA) to improve the efficiency of analysis techniques. The objective of the FA2IR project is to develop enhanced AI-based methods for microelectronic failure analysis and to get FA databases AI-ready. The project ‘s approach to the database landscape is unique due to the implementation of the FAIR-data principle (Findable, Accessible, Interoperable, Reusable) and new methods for AI-based database search. The novel and improved data landscapes will ensure data standardization and interoperability of databases by leveraging expert knowledge stored in FA-ontology. Therefore, FA ontology and a universally accepted data format will be employed to enable companies and suppliers to effortlessly exchange data while ensuring effective protection of intellectual property and confidential information. Linking databases within the value chain of semiconductor development and production process will enable a deeper understanding of failure modes, including failure prediction in the design and qualification phase of new products based on existing knowledge from comparable older products. Additionally, improved AI-based methods for FA, such as image and measurement data analysis, text classification, etc., will be developed. The average analysis time will decrease, and a higher level of data standardization will be achieved based on the planed project results.

The ILH is involved in the exploration and development of novel AI-based correlative data analysis techniques, including the concept of multimodal models. Specific application scenarios encompass multi-modal Machine Learning (ML) models with Time Domain Reflectometry (TDR) or Temperature Sensitive Electrical Parameter (TSEP) signal modalities, enabling more precise and non-destructive failure detection in wide-bandgap semiconductor devices. Additionally, AI-ready database infrastructures and standardized ontologies for the field of the failure analysis in semiconductor devices will be established through collaboration with national and international partners. The goal is to enable automation of data processing and integration, making failure analysis more efficient and accurate.

Germany: Infineon (including Uni Klagenfurth), Bosch, Zeiss, Fraunhofer IWMS, PVA TePLA, Matworks, University of Stuttgart

Sweden: RISE, Ericsson,  Gimic,  Alstom

Netherlands: NXP, Maser